verge-rss

Prime Video will let you summon AI to recap what you’re watching

Image: Amazon

Amazon’s Prime Video is getting a new generative AI-powered feature to help catch you up on a show. The new tool, called X-Ray Recaps, can create text summaries of “of full seasons of TV shows, single episodes, and even pieces of episodes,” the company says in a blog post.
X-Ray Recaps will be accessible from the detail page of a show or in X-Ray while you’re watching something. The tool “analyzes various video segments, combined with subtitles or dialogue, to generate detailed descriptions of key events, places, times, and conversations,” Amazon says. Amazon has also applied “guardrails” to help the feature avoid sharing spoilers and to keep summaries concise.

Image: Amazon

X-Ray Recaps, which is are beta, are coming to Fire TV devices starting today, with support for “additional devices” available by the end of this year, Amazon says. The feature, at launch, will work with all Amazon MGM Studios Original series.

Image: Amazon

X-Ray Recaps will be accessible from the detail page of a show or in X-Ray while you’re watching something. The tool “analyzes various video segments, combined with subtitles or dialogue, to generate detailed descriptions of key events, places, times, and conversations,” Amazon says. Amazon has also applied “guardrails” to help the feature avoid sharing spoilers and to keep summaries concise.

Image: Amazon

Elon Musk’s PAC admits $1 million voter giveaways aren’t ‘random’

Image: Cath Virginia / The Verge, Getty Images

A representative of Elon Musk’s America PAC said the winners of its $1 million voter sweepstakes aren’t randomly selected. The people who are awarded the checks are chosen to be paid “spokespeople” for the pro-Trump PAC, lawyer Chris Gober testified in a Philadelphia court on Monday, as reported by the Associated Press.
Philadelphia district attorney Larry Krasner sued Musk and his PAC over the contest, alleging it’s an illegal lottery that violates Pennsylvania state laws. “Though Musk says that a winner’s selection is ‘random,’ that appears false because multiple winners that have been selected are individuals who have shown up at rallies in Pennsylvania,” the suit read. America PAC’s response confirms Krasner’s allegation that the winners of the contest aren’t selected at random — but this is presented as proof that the sweepstakes isn’t a lottery at all.
“The $1 million recipients are not chosen by chance,” Gober — who CNBC identified as the former treasurer of the PAC — said during a Monday hearing, adding that the next two recipients will be from Arizona and Michigan. “We know exactly who will be announced as the $1 million recipient today and tomorrow.”

ELON MUSK: “I have a surprise for you. We are going to be awarding $1 MILLION to people who have signed the petition every day, from now until the election.”SIGN America PAC’s petition in support of Free Speech & The Right to Bear Arms: https://t.co/TMeyWUhbrH pic.twitter.com/QcgePZrhMr— America (@america) October 20, 2024

This contradicts Musk’s own claim that the awards were granted “randomly.” Musk announced the contest at a rally in Harrisburg, Pennsylvania, in late October. The event, one of several Musk held for Trump in the battleground state, was only open to people who had signed an America PAC petition saying they support the First and Second Amendments — a petition that critics quickly noted could only be signed by registered voters in Pennsylvania, a key swing state. America PAC eventually expanded it to six other swing states.
But the petition’s language doesn’t say anything about winners being selected randomly. Instead, it reads:
Each day through Nov. 5, one petition signer from either PA, GA, NV, AZ, MI, WI, or NC will earn $1,000,000.
And Grober testified that Musk’s use of the word randomly “is used interchangeably with the words ‘by chance.’” John Summers, the attorney representing Krasner’s office, described Grober’s comments as a “flat-out admission of liability,” CNBC reports.
Per the AP, more than one million people across the seven states have signed the petition thus far. Krasner, the Philadelphia DA, raised questions about how the PAC may continue to use those users’ data. “They were scammed for their information,” Krasner said during the Monday hearing. “It has almost unlimited use.”

Image: Cath Virginia / The Verge, Getty Images

Philadelphia district attorney Larry Krasner sued Musk and his PAC over the contest, alleging it’s an illegal lottery that violates Pennsylvania state laws. “Though Musk says that a winner’s selection is ‘random,’ that appears false because multiple winners that have been selected are individuals who have shown up at rallies in Pennsylvania,” the suit read. America PAC’s response confirms Krasner’s allegation that the winners of the contest aren’t selected at random — but this is presented as proof that the sweepstakes isn’t a lottery at all.

“The $1 million recipients are not chosen by chance,” Gober — who CNBC identified as the former treasurer of the PAC — said during a Monday hearing, adding that the next two recipients will be from Arizona and Michigan. “We know exactly who will be announced as the $1 million recipient today and tomorrow.”

ELON MUSK: “I have a surprise for you. We are going to be awarding $1 MILLION to people who have signed the petition every day, from now until the election.”

SIGN America PAC’s petition in support of Free Speech & The Right to Bear Arms: https://t.co/TMeyWUhbrH pic.twitter.com/QcgePZrhMr

— America (@america) October 20, 2024

But the petition’s language doesn’t say anything about winners being selected randomly. Instead, it reads:

Each day through Nov. 5, one petition signer from either PA, GA, NV, AZ, MI, WI, or NC will earn $1,000,000.

And Grober testified that Musk’s use of the word randomly “is used interchangeably with the words ‘by chance.’” John Summers, the attorney representing Krasner’s office, described Grober’s comments as a “flat-out admission of liability,” CNBC reports.

Per the AP, more than one million people across the seven states have signed the petition thus far. Krasner, the Philadelphia DA, raised questions about how the PAC may continue to use those users’ data. “They were scammed for their information,” Krasner said during the Monday hearing. “It has almost unlimited use.”

Amazon’s Echo Spot alarm clock is on sale with a free color smart bulb

The Echo Spot is a smart alarm clock / speaker hybrid that’s $5 shy of its all-time low price. | Photo by Jennifer Pattison Tuohy / The Verge

The clocks have fallen back an hour for many of us, and if it all feels a bit disorienting, you’re not alone. Thankfully, Amazon’s versatile Echo Spot is on sale to help you adjust. Normally $79.99, right now you can buy the smart alarm clock at Amazon in black, blue, or white with a free Globe Electric smart bulb for $49.99, which equates to $41.99 in savings. That’s just $5 shy of the all-time low we saw during Amazon’s recent Prime Day event.

Amazon’s speedy smart speaker can be set up so that it gently wakes you up with music instead of typical alarm clock sounds, which can be jarring. The Spot is also a lot more useful than your run-of-the-mill clock, as it offers a customizable 2.83-inch screen that displays helpful info (including the weather and music playback). However, unlike Amazon’s larger smart displays, the latest Spot doesn’t push on-screen ads and lacks a camera, so there’s less of a privacy concern.
What makes the Spot particularly useful, though, is that it functions as an inexpensive Alexa speaker. That means you can use it to perform all kinds of tasks, from setting reminders to playing podcasts, audiobooks, and music. You can also use it to control other smart home devices with just your voice, including lights and smart thermostats. That’ll come in handy as the days get colder and darker — after all, no one wants to leave the warmth of their bed just to hit the lights.
Read our Echo Spot review.

A few other ways to save

The Fitbit Charge 6 is down to an all-time low of $94.95 ($65 off) at Macy’s in multiple colors. The high-end tracker offers all the standard health and fitness features you’d expect, along with sensors for tracking things like stress. However, unlike the prior model, it also supports Google Wallet and turn-by-turn navigations via Google Maps. It even offers a color OLED display and can broadcast your heart rate on select Bluetooth gym equipment.

Samsung’s Music Frame is down to a new low price of around $248 ($150 off) at Amazon and Best Buy. The mountable, 12.9 x 12.9-inch picture frame functions as a wired Bluetooth speaker that works over Wi-Fi, one with support for Dolby Atmos, Alexa, and Samsung’s Bixby voice assistant. You can insert physical photographs or art prints, too, and if you own one of Samsung’s new TVs or soundbars, you can even pair them with the Frame for a surround sound experience.
You can buy Amazon’s new Fire HD 8 at Amazon and Best Buy with ads and 32GB of storage for a new low of $54.99, down from $99.99. Unlike its predecessor, the new budget-friendly tablet offers 3GB of RAM (up from 2GB) and generative AI-powered tools, including a feature that can polish your writing and another that summarizes webpages. Otherwise, though, it’s pretty similar to its predecessor, with an identical eight-inch 1280 x 800 LCD display and up to 13 hours of battery life.

The Echo Spot is a smart alarm clock / speaker hybrid that’s $5 shy of its all-time low price. | Photo by Jennifer Pattison Tuohy / The Verge

The clocks have fallen back an hour for many of us, and if it all feels a bit disorienting, you’re not alone. Thankfully, Amazon’s versatile Echo Spot is on sale to help you adjust. Normally $79.99, right now you can buy the smart alarm clock at Amazon in black, blue, or white with a free Globe Electric smart bulb for $49.99, which equates to $41.99 in savings. That’s just $5 shy of the all-time low we saw during Amazon’s recent Prime Day event.

What makes the Spot particularly useful, though, is that it functions as an inexpensive Alexa speaker. That means you can use it to perform all kinds of tasks, from setting reminders to playing podcasts, audiobooks, and music. You can also use it to control other smart home devices with just your voice, including lights and smart thermostats. That’ll come in handy as the days get colder and darker — after all, no one wants to leave the warmth of their bed just to hit the lights.

Read our Echo Spot review.

A few other ways to save

The Fitbit Charge 6 is down to an all-time low of $94.95 ($65 off) at Macy’s in multiple colors. The high-end tracker offers all the standard health and fitness features you’d expect, along with sensors for tracking things like stress. However, unlike the prior model, it also supports Google Wallet and turn-by-turn navigations via Google Maps. It even offers a color OLED display and can broadcast your heart rate on select Bluetooth gym equipment.

Samsung’s Music Frame is down to a new low price of around $248 ($150 off) at Amazon and Best Buy. The mountable, 12.9 x 12.9-inch picture frame functions as a wired Bluetooth speaker that works over Wi-Fi, one with support for Dolby Atmos, Alexa, and Samsung’s Bixby voice assistant. You can insert physical photographs or art prints, too, and if you own one of Samsung’s new TVs or soundbars, you can even pair them with the Frame for a surround sound experience.
You can buy Amazon’s new Fire HD 8 at Amazon and Best Buy with ads and 32GB of storage for a new low of $54.99, down from $99.99. Unlike its predecessor, the new budget-friendly tablet offers 3GB of RAM (up from 2GB) and generative AI-powered tools, including a feature that can polish your writing and another that summarizes webpages. Otherwise, though, it’s pretty similar to its predecessor, with an identical eight-inch 1280 x 800 LCD display and up to 13 hours of battery life.

Can Philadelphia’s ballot counters outrun election lies?

The machines that process mail-in ballots help count thousands of votes in a day — and Philadelphia officials know that every second matters. The room where it happens has metal beams and harsh overhead lighting. Paper whizzes through conveyor belts on large gears near tall, human-sized cages with keypad locks.
Though it resembles one, this is not a factory. It’s Philadelphia’s mail-in ballot-counting facility, where somewhere around 200,000 votes are expected to be tallied beginning on Election Day. The longer that tally takes, the more misinformation could seep into a deep well of paranoia and distrust over the democratic process — one that overflowed four years ago in a violent attack on the US Capitol.
The stakes, you could say, are high.
Pennsylvania’s 19 electoral votes could decide whether former President Donald Trump or Vice President Kamala Harris wins the 2024 presidential election. Many of the state’s ballots have already been cast through the mail. Yet Pennsylvania’s laws prohibit even beginning to process mail-in ballots until 7AM on Election Day. The result can be a serious delay in reporting election results — in 2020, The Associated Press didn’t call Pennsylvania as having been won by Joe Biden until four days after Election Day.
The state’s solution is a downright industrial ballot-counting process, which elected officials invited reporters to preview (using test ballots for demonstration purposes) in late October. It’s a highly regimented process that takes place in a sprawling warehouse in northeast Philadelphia, filled with the sounds of whirring ballot-sorting machines and the constant rifling of paper. On Election Day, workers will open hundreds of thousands of mailed ballots and feed them into machines that read and count them, keeping a careful eye on monitors flagging any irregularities. The scanned ballots will also be watched by election observers from each political party. “We do it right,” says Philadelphia City Commissioners chair Omar Sabir, a Democrat.
“The more people hear things, unfortunately, the more inclined they are to believe them.”
The city’s press tour is part of a broader effort to educate voters and reassure them that voting is safe, secure, and trustworthy. It’s an attempted bulwark against false claims about ballots being inaccurately tallied, flipped, or destroyed to skew election results.
Lisa Deeley, Democratic vice chair of the Philadelphia City Commissioners, says she doesn’t expect as long of a delay as 2020. Mail-in voting was an unusually popular option that year due to the ongoing covid-19 pandemic. But when conspiracies can ricochet across social media in seconds, every hour counts. “The more people hear things, unfortunately, the more inclined they are to believe them,” she says.

As Deeley explains, election workers are “starting from brick one” on Election Day. That means not just tallying who votes for whom, but reviewing the signatures on sealed envelopes, removing them from their secrecy sleeves, and flattening the ballots themselves. Commissioners have been “begging” for reforms to this process, Deeley says. Absent these changes, they’re left with technical and procedural solutions like buying new equipment and relying on more experienced election workers — shaving time off the clock any way they can.
“We know that the eyes of the world are going to be on Philadelphia,” City Commissioner Seth Bluestein, a Republican, told reporters gathered in the warehouse. “We are going to run the safest, most secure election in Philadelphia history.”

The process of tallying ballots — as I and other reporters see, shuffling behind Sabir around the 360,000-square-foot room — starts with what looks like an oversized Xerox machine. It feeds ballots in sealed envelopes into a conveyor belt on two gears and spits them out across a long track, sorting them into different slots based on ward and division. The machine scans barcodes on the envelopes, each one linked to a registered voter’s ID to mark the ballots as “received” so one voter can’t send multiple votes. If it was mailed without a signature or not placed in its included secrecy envelope, it’s set aside and added to a list that voters can check, letting them correct the problem with a replacement ballot.
This is as far as workers can get before Election Day, so the sorted envelopes go into secure storage until the morning of November 5th. Then, at the crack of dawn, the count begins. Twenty-two envelope extractors, built around desks where workers will help separate the envelopes from their contents once opened, run about 1,000 envelopes each per hour. Four rapid slicing machines open the yellow secrecy envelopes inside those envelopes at a rate of about 10,000 per hour. Workers remove the ballots from the now-opened secrecy envelopes — and since this process is separated from when the ballots are removed from their outer envelopes, it ensures votes stay anonymous. Now patted flat, the ballots reach the step this whole process is building up to: the count.

An election scanner is basically a gigantic Scantron machine, with a stretched-out metal S-shape that ballots glide through as the machine reads the marks voters have made. This warehouse has eight high-speed scanners, each one expected to check about 2,500 ballots per hour. (Four additional slower scanners can read 1,000 per hour.) Some ballots can’t be read — if they’ve been marked with a light-colored pen or had mistakes erased with Wite-Out, for instance. A staff of nonpartisan civil servants review these and mark the voter’s choices onto replacement ballots, which can then be scanned.
Election observers — who are selected by each party — will watch screens showing the ballots to help ensure everything is adjudicated fairly. Finally, the processed ballots go into another locked storage area. They will ultimately be kept in long-term storage for the 22 months mandated by law — just in case they’re needed for a recount.

As this count is taking place, an opposing process will be spinning up: a disinformation apparatus that aims to convince voters the election is being rigged.
In 2020, this process coalesced into the “Stop the Steal” election denial movement, culminating in an attempt to overturn the election of President Joe Biden by force. In 2024, it’s already gotten started. A group of Republicans including House Rep. Scott Perry (R-PA), who voted to sustain objections to the 2020 election results, sued Pennsylvania’s state government, demanding military and overseas ballots be set aside because of what election experts call unfounded doubts about the process. (It was also recently tossed by a judge.) Election deniers have gained seats on important state and local bodies that could give them leverage over election certification. And online, Trump mega-donor Elon Musk has set up an X community for reporting “voter fraud and irregularities,” which has already filled up with unfounded claims.
For Philadelphia’s City Commissioners, misinformation is personal. Sabir smiles as he relates one of the “craziest” conspiracy theories: a blog post that claimed he was personally taking ballots to a mobster in Atlantic City, New Jersey, to be destroyed. (It’s not clear why the mobster needs them trucked out more than 60 miles just to be shredded.) “We’re not doing crazy stuff. We’re just trying to come to do our job for the American people,” Sabir says.

The threats aren’t always amusing. Bluestein told The Verge that, while ballots were being counted in 2020, he received antisemitic threats. The harassment got so bad that Bluestein had police protection at his house the week of the election.
So far, “the heat is down” in 2024, he says. But election officials around the country are still on high alert. In Maricopa County in Arizona, another swing state, an official recently said security will be available to escort election workers to their cars. The state is also preparing for cutting-edge risks like infiltration by artificial intelligence scams, a scenario the staff roleplayed last year.
Social media platforms “are not doing as good a job as they did in 2020” with combating misinformation
Bluestein himself is trying to spot and call out false information online. In one case, he debunked an allegation — shared on X by Musk — that a nonprofit offering services to low-income and houseless individuals harvested thousands of mail-in ballots from one address. (Bluestein says “fewer than 150 ballots” were mailed there in 2020.)
His active role online is partially because he feels that social media platforms “are not doing as good a job as they did in 2020” with combating misinformation. Four years ago, platforms were on high alert for false claims, even if they often failed to enforce their policies effectively. In 2024, the situation is different. Under pressure from Trump and his allies to take a more hands-off role when it comes to election misinformation, many tech companies have relaxed the policies they had in place last time around. Meta and YouTube both rolled back rules against false claims that the 2020 election was stolen, and both Meta and X have made it more difficult for researchers on their platforms to access data used to monitor emerging threats.
At the same time, both Bluestein and Sabir say they haven’t yet seen the same level of targeted harassment and threats. Bluestein says broader misinformation claims are circulating, but he hasn’t found as many claims that single out specific officials or rank-and-file workers. Despite harassment in 2020, he says Philadelphia had no problem recruiting poll workers or staffers at the warehouse. “I think everyone understands the importance of this work, and they understand that while there could be risks associated with it, they’re all signing up to do the job.”

Misinformation often picks at tensions that already exist. Black Americans, for example, are already a group commonly targeted by disenfranchisement efforts, which Sabir says results in “misconceptions about ‘my vote doesn’t count.’” Add disinformation to that, and Sabir says it drives a notion through this community of “What am I doing? Why am I wasting my time?”
Bluestein has found that he can persuade voters through one-on-one conversations. But during that time, false claims can reach millions of voters online. “When you scale that up to build trust, it’s a lot harder,” Bluestein says. “But when you really tell people the facts and show them, they will have more faith.”
This year, election deniers are using “administrative tactics” to suppress votes
While election officials try to persuade skeptics, election deniers have increased their attacks against the administration of the voting process itself. In 2020, election deniers “used violent rhetoric as the means to suppress the vote and make it harder for folks to vote, or make the voting process seem scary and intimidating,” says Deborah Hinchey, Pennsylvania state director for the nonpartisan nonprofit All Voting is Local. This year, she is seeing election deniers use “administrative tactics to do the same thing — to suppress the vote, to make it seem an intimidating and overwhelming process, and to make folks feel like their vote may not be counted.”
But Hinchey says those efforts will fail. In 2020, Trump lawyers and other allies brought numerous cases to change the election results after the fact. Those suits invariably fizzled, and some of the lawyers who filed them have been sanctioned or disbarred. “The analysis now seems to be, ‘Well, then let’s go directly for the votes themselves, and discredit certain kinds of voters and make it seem like certain people are voting that are not, so that we can then attack all votes.’”
So far, these attempts largely haven’t panned out. While right-wing activists throughout Pennsylvania have sought to challenge voter registrations, they’ve proven unsuccessful or identified inactive voters election officials already knew about.
As for trust, a September Spotlight PA poll by MassINC Polling Group found that 63 percent of respondents were very or somewhat confident that votes in the presidential race would be counted accurately and fairly nationally. But voters had far more confidence in how elections in their own counties would be administered — 78 percent expressed confidence in the results.

Organizers are seeing more people wanting to get involved in the process of democracy, and that participation can help quell election fears, says Susan Gobreski, president of the League of Women Voters of Philadelphia. While hearing about election skeptics getting involved in the process might raise some red flags, Gobreski says it’s important to remember that “most people are actually acting in good faith.”
Arming the public and the press with trustworthy information is a smart move, says Hinchey. “You can’t dispel all bad information with good information, but you can make sure that organizations and the press have a really good understanding of how elections are actually functioning in Pennsylvania,” she says. Gobreski encourages voters to ask questions but also to “be prepared to listen to the answers.”
Ultimately, Hinchey adds, most voters are just looking for reliable information. “The average Pennsylvania voter is looking for the facts of the situation, and may take in the falseness, but when presented with facts, is going to accept that as reality.”

There’s one final option on the table for ballot tallies: a hand count. It’s common to audit samples of ballots by hand and compare them to machine results, confirming the machines are working properly. (Election officials also do preelection testing of equipment to make sure they’re properly calibrated, often on livestreams.) But in states like Georgia, election skeptics have — so far, unsuccessfully — pushed for full hand counts of every ballot. That’s a recipe for mistakes and delays.
Hand-counting has an important role in auditing elections, says Pennsylvania Secretary of the Commonwealth Al Schmidt, a Republican. “But if you’re just counting by hand, you don’t have anything to compare it against. So when people do significant numbers of hand counting, that’s where you see more errors.” Compared to machine counts, it’s also a glacial process. “If people are unhappy with how long they wait now, imagine how long” it would take without the machinery to get results, says Deeley. “It’s [like] going to Nabisco and having them make all the cookies by scratch.”
On Election Day, the machines in Philadelphia’s warehouse will flip on, rifling through thousands and thousands of envelopes, slicing and scanning. Signatures will be checked and folded. Paper will be flattened. The work of democracy will run through machines and careful human hands. It’s a tedious process, but it’s also one that’s at the very heart of the American experiment. Each ballot counted is one step closer to determining if Pennsylvania will be colored in red or blue on TV screens across the country — and possibly determining the next president. And although every minute after polls close is another minute for spreading doubt in America’s electoral system, Philadelphia’s officials are resolute.
“Philadelphia is the birthplace of democracy,” says Sabir. “I’ll be damned if democracy dies here.”

The machines that process mail-in ballots help count thousands of votes in a day — and Philadelphia officials know that every second matters.

The room where it happens has metal beams and harsh overhead lighting. Paper whizzes through conveyor belts on large gears near tall, human-sized cages with keypad locks.

Though it resembles one, this is not a factory. It’s Philadelphia’s mail-in ballot-counting facility, where somewhere around 200,000 votes are expected to be tallied beginning on Election Day. The longer that tally takes, the more misinformation could seep into a deep well of paranoia and distrust over the democratic process — one that overflowed four years ago in a violent attack on the US Capitol.

The stakes, you could say, are high.

Pennsylvania’s 19 electoral votes could decide whether former President Donald Trump or Vice President Kamala Harris wins the 2024 presidential election. Many of the state’s ballots have already been cast through the mail. Yet Pennsylvania’s laws prohibit even beginning to process mail-in ballots until 7AM on Election Day. The result can be a serious delay in reporting election results — in 2020, The Associated Press didn’t call Pennsylvania as having been won by Joe Biden until four days after Election Day.

The state’s solution is a downright industrial ballot-counting process, which elected officials invited reporters to preview (using test ballots for demonstration purposes) in late October. It’s a highly regimented process that takes place in a sprawling warehouse in northeast Philadelphia, filled with the sounds of whirring ballot-sorting machines and the constant rifling of paper. On Election Day, workers will open hundreds of thousands of mailed ballots and feed them into machines that read and count them, keeping a careful eye on monitors flagging any irregularities. The scanned ballots will also be watched by election observers from each political party. “We do it right,” says Philadelphia City Commissioners chair Omar Sabir, a Democrat.

“The more people hear things, unfortunately, the more inclined they are to believe them.”

The city’s press tour is part of a broader effort to educate voters and reassure them that voting is safe, secure, and trustworthy. It’s an attempted bulwark against false claims about ballots being inaccurately tallied, flipped, or destroyed to skew election results.

Lisa Deeley, Democratic vice chair of the Philadelphia City Commissioners, says she doesn’t expect as long of a delay as 2020. Mail-in voting was an unusually popular option that year due to the ongoing covid-19 pandemic. But when conspiracies can ricochet across social media in seconds, every hour counts. “The more people hear things, unfortunately, the more inclined they are to believe them,” she says.

“We know that the eyes of the world are going to be on Philadelphia,” City Commissioner Seth Bluestein, a Republican, told reporters gathered in the warehouse. “We are going to run the safest, most secure election in Philadelphia history.”

This is as far as workers can get before Election Day, so the sorted envelopes go into secure storage until the morning of November 5th. Then, at the crack of dawn, the count begins. Twenty-two envelope extractors, built around desks where workers will help separate the envelopes from their contents once opened, run about 1,000 envelopes each per hour. Four rapid slicing machines open the yellow secrecy envelopes inside those envelopes at a rate of about 10,000 per hour. Workers remove the ballots from the now-opened secrecy envelopes — and since this process is separated from when the ballots are removed from their outer envelopes, it ensures votes stay anonymous. Now patted flat, the ballots reach the step this whole process is building up to: the count.

Election observers — who are selected by each party — will watch screens showing the ballots to help ensure everything is adjudicated fairly. Finally, the processed ballots go into another locked storage area. They will ultimately be kept in long-term storage for the 22 months mandated by law — just in case they’re needed for a recount.

As this count is taking place, an opposing process will be spinning up: a disinformation apparatus that aims to convince voters the election is being rigged.

In 2020, this process coalesced into the “Stop the Steal” election denial movement, culminating in an attempt to overturn the election of President Joe Biden by force. In 2024, it’s already gotten started. A group of Republicans including House Rep. Scott Perry (R-PA), who voted to sustain objections to the 2020 election results, sued Pennsylvania’s state government, demanding military and overseas ballots be set aside because of what election experts call unfounded doubts about the process. (It was also recently tossed by a judge.) Election deniers have gained seats on important state and local bodies that could give them leverage over election certification. And online, Trump mega-donor Elon Musk has set up an X community for reporting “voter fraud and irregularities,” which has already filled up with unfounded claims.

For Philadelphia’s City Commissioners, misinformation is personal. Sabir smiles as he relates one of the “craziest” conspiracy theories: a blog post that claimed he was personally taking ballots to a mobster in Atlantic City, New Jersey, to be destroyed. (It’s not clear why the mobster needs them trucked out more than 60 miles just to be shredded.) “We’re not doing crazy stuff. We’re just trying to come to do our job for the American people,” Sabir says.

The threats aren’t always amusing. Bluestein told The Verge that, while ballots were being counted in 2020, he received antisemitic threats. The harassment got so bad that Bluestein had police protection at his house the week of the election.

So far, “the heat is down” in 2024, he says. But election officials around the country are still on high alert. In Maricopa County in Arizona, another swing state, an official recently said security will be available to escort election workers to their cars. The state is also preparing for cutting-edge risks like infiltration by artificial intelligence scams, a scenario the staff roleplayed last year.

Social media platforms “are not doing as good a job as they did in 2020” with combating misinformation

Bluestein himself is trying to spot and call out false information online. In one case, he debunked an allegation — shared on X by Musk — that a nonprofit offering services to low-income and houseless individuals harvested thousands of mail-in ballots from one address. (Bluestein says “fewer than 150 ballots” were mailed there in 2020.)

His active role online is partially because he feels that social media platforms “are not doing as good a job as they did in 2020” with combating misinformation. Four years ago, platforms were on high alert for false claims, even if they often failed to enforce their policies effectively. In 2024, the situation is different. Under pressure from Trump and his allies to take a more hands-off role when it comes to election misinformation, many tech companies have relaxed the policies they had in place last time around. Meta and YouTube both rolled back rules against false claims that the 2020 election was stolen, and both Meta and X have made it more difficult for researchers on their platforms to access data used to monitor emerging threats.

At the same time, both Bluestein and Sabir say they haven’t yet seen the same level of targeted harassment and threats. Bluestein says broader misinformation claims are circulating, but he hasn’t found as many claims that single out specific officials or rank-and-file workers. Despite harassment in 2020, he says Philadelphia had no problem recruiting poll workers or staffers at the warehouse. “I think everyone understands the importance of this work, and they understand that while there could be risks associated with it, they’re all signing up to do the job.”

Bluestein has found that he can persuade voters through one-on-one conversations. But during that time, false claims can reach millions of voters online. “When you scale that up to build trust, it’s a lot harder,” Bluestein says. “But when you really tell people the facts and show them, they will have more faith.”

This year, election deniers are using “administrative tactics” to suppress votes

While election officials try to persuade skeptics, election deniers have increased their attacks against the administration of the voting process itself. In 2020, election deniers “used violent rhetoric as the means to suppress the vote and make it harder for folks to vote, or make the voting process seem scary and intimidating,” says Deborah Hinchey, Pennsylvania state director for the nonpartisan nonprofit All Voting is Local. This year, she is seeing election deniers use “administrative tactics to do the same thing — to suppress the vote, to make it seem an intimidating and overwhelming process, and to make folks feel like their vote may not be counted.”

But Hinchey says those efforts will fail. In 2020, Trump lawyers and other allies brought numerous cases to change the election results after the fact. Those suits invariably fizzled, and some of the lawyers who filed them have been sanctioned or disbarred. “The analysis now seems to be, ‘Well, then let’s go directly for the votes themselves, and discredit certain kinds of voters and make it seem like certain people are voting that are not, so that we can then attack all votes.’”

So far, these attempts largely haven’t panned out. While right-wing activists throughout Pennsylvania have sought to challenge voter registrations, they’ve proven unsuccessful or identified inactive voters election officials already knew about.

As for trust, a September Spotlight PA poll by MassINC Polling Group found that 63 percent of respondents were very or somewhat confident that votes in the presidential race would be counted accurately and fairly nationally. But voters had far more confidence in how elections in their own counties would be administered — 78 percent expressed confidence in the results.

Arming the public and the press with trustworthy information is a smart move, says Hinchey. “You can’t dispel all bad information with good information, but you can make sure that organizations and the press have a really good understanding of how elections are actually functioning in Pennsylvania,” she says. Gobreski encourages voters to ask questions but also to “be prepared to listen to the answers.”

Ultimately, Hinchey adds, most voters are just looking for reliable information. “The average Pennsylvania voter is looking for the facts of the situation, and may take in the falseness, but when presented with facts, is going to accept that as reality.”

Hand-counting has an important role in auditing elections, says Pennsylvania Secretary of the Commonwealth Al Schmidt, a Republican. “But if you’re just counting by hand, you don’t have anything to compare it against. So when people do significant numbers of hand counting, that’s where you see more errors.” Compared to machine counts, it’s also a glacial process. “If people are unhappy with how long they wait now, imagine how long” it would take without the machinery to get results, says Deeley. “It’s [like] going to Nabisco and having them make all the cookies by scratch.”

On Election Day, the machines in Philadelphia’s warehouse will flip on, rifling through thousands and thousands of envelopes, slicing and scanning. Signatures will be checked and folded. Paper will be flattened. The work of democracy will run through machines and careful human hands. It’s a tedious process, but it’s also one that’s at the very heart of the American experiment. Each ballot counted is one step closer to determining if Pennsylvania will be colored in red or blue on TV screens across the country — and possibly determining the next president. And although every minute after polls close is another minute for spreading doubt in America’s electoral system, Philadelphia’s officials are resolute.

“Philadelphia is the birthplace of democracy,” says Sabir. “I’ll be damned if democracy dies here.”

PS5 Pro’s actual specs and the launch day list of upgraded games revealed

Image: Sony

Ahead of the PS5 Pro launch later this week, we have confirmation of a few more official specifications for the most powerful version of the PS5 yet. Digital Foundry has shared details from the manual included with its PS5 Pro review unit, confirming the console is powered by an AMD Ryzen Zen 2 8-core/16-thread processor with an RDNA-based graphics engine producing 16.7TF of GPU compute performance compared to the base PS5’s 10.23TF.
As we noted in March, initial leaked documents showed a much higher figure for the PS5 Pro, mostly due to a change in AMD’s RDNA architecture.

The manual also reveals that in addition to 16GB of GDDR6 memory, the PS5 Pro will have 2GB of slower DDR5 memory for handling console tasks not related to gaming, plus a 2TB custom SSD. At 390W, its maximum power draw will also be more than the 340W used by the slim and regular versions of the PS5 Digital Edition, but the console won’t be drawing that much power at all times.
In a post shared to the official PlayStation blog today, Sony also revealed a list of 55 games that will have PS5 Pro enhancements when the $699.99 console launches on November 7th. The list includes several titles already revealed to be getting updates but adds several others, such as Call of Duty: Black Ops 6 and Star Wars: Outlaws.

Alan Wake 2
Albatroz
Apex Legends
Arma Reforger
Assassin’s Creed Mirage
Baldur’s Gate 3
Call of Duty: Black Ops 6
EA Sports College Football 25
Dead Island 2
Demon’s Souls
Diablo IV
Dragon Age: The Veilguard
Dragon’s Dogma 2
Dying Light 2 Reloaded Edition
EA Sports FC 25
Enlisted
F1 24
Final Fantasy VII Rebirth
Fortnite
God of War Ragnarök
Hogwarts Legacy
Horizon Forbidden West
Horizon Zero Dawn Remastered
Kayak VR: Mirage
Lies of P
Madden NFL 25
Marvel’s Spider-Man Remastered
Marvel’s Spider-Man: Miles Morales
Marvel’s Spider-Man 2
Naraka: Bladepoint
NBA2K 25
No Man’s Sky
Palworld
Paladin’s Passage
Planet Coaster 2
Professional Spirits Baseball 2024-2025
Ratchet & Clank: Rift Apart
Resident Evil 4
Resident Evil Village
Rise of the Ronin
Rogue Flight
Star Wars: Jedi Survivor
Star Wars: Outlaws
Stellar Blade
Test Drive Unlimited: Solar Crown
The Callisto Protocol
The Crew Motorfest
The Finals
The First Descendant
The Last of Us Part I
The Last of Us Part II Remastered
Until Dawn
War Thunder
Warframe
World of Warships: Legends

Image: Sony

Ahead of the PS5 Pro launch later this week, we have confirmation of a few more official specifications for the most powerful version of the PS5 yet. Digital Foundry has shared details from the manual included with its PS5 Pro review unit, confirming the console is powered by an AMD Ryzen Zen 2 8-core/16-thread processor with an RDNA-based graphics engine producing 16.7TF of GPU compute performance compared to the base PS5’s 10.23TF.

As we noted in March, initial leaked documents showed a much higher figure for the PS5 Pro, mostly due to a change in AMD’s RDNA architecture.

In a post shared to the official PlayStation blog today, Sony also revealed a list of 55 games that will have PS5 Pro enhancements when the $699.99 console launches on November 7th. The list includes several titles already revealed to be getting updates but adds several others, such as Call of Duty: Black Ops 6 and Star Wars: Outlaws.

Alan Wake 2
Albatroz
Apex Legends
Arma Reforger
Assassin’s Creed Mirage
Baldur’s Gate 3
Call of Duty: Black Ops 6
EA Sports College Football 25
Dead Island 2
Demon’s Souls
Diablo IV
Dragon Age: The Veilguard
Dragon’s Dogma 2
Dying Light 2 Reloaded Edition
EA Sports FC 25
Enlisted
F1 24
Final Fantasy VII Rebirth
Fortnite
God of War Ragnarök
Hogwarts Legacy
Horizon Forbidden West
Horizon Zero Dawn Remastered
Kayak VR: Mirage
Lies of P
Madden NFL 25
Marvel’s Spider-Man Remastered
Marvel’s Spider-Man: Miles Morales
Marvel’s Spider-Man 2
Naraka: Bladepoint
NBA2K 25
No Man’s Sky
Palworld
Paladin’s Passage
Planet Coaster 2
Professional Spirits Baseball 2024-2025
Ratchet & Clank: Rift Apart
Resident Evil 4
Resident Evil Village
Rise of the Ronin
Rogue Flight
Star Wars: Jedi Survivor
Star Wars: Outlaws
Stellar Blade
Test Drive Unlimited: Solar Crown
The Callisto Protocol
The Crew Motorfest
The Finals
The First Descendant
The Last of Us Part I
The Last of Us Part II Remastered
Until Dawn
War Thunder
Warframe
World of Warships: Legends

X is allowing people you’ve blocked to see your posts

Illustration: The Verge

This weekend, X, the platform formerly known as Twitter, announced it’s “starting to launch” a controversial change to how blocking works on its platform. Company owner Elon Musk first revealed the change in September, which will allow people you’ve blocked to continue to see your posts, and, as noted by TechCrunch, your following and followers lists.
Musk claimed that stopping people from seeing your public posts “makes no sense,” but due to a post-Musk change that stops logged-out users from scrolling even a public profile, this could make it easier for blocked users to continue harassing someone.
In October, X’s engineering account argued that people who block others could say harmful things about the blocked person who wouldn’t know about it and that the change allows for “greater transparency.” The change will still disable blocked users from following, liking, replying, reposting, DMing, or interacting with posts from the blocking party.
But as Tracy Chou, who launched the anti-harassment tool Block Party, said in a post on X, “…the point is that friction matters!! making it easy for a creeper to creep is not a good thing!!”
Another issue is whether the updated block function aligns with the rules in app stores. Apple’s guidelines say that apps with user-generated content must have “The ability to block abusive users from the service.”
Google’s guidelines are more specific:

UGC features that enable 1:1 user interaction with specific users (for example, direct messaging, tagging, mentioning, etc.) must provide an in-app functionality for blocking users.
Apps that provide access to publicly accessible UGC, such as social networking apps and blogger apps, must implement in-app functionality to report users and content, and to block users.

We’re starting to launch the block function update https://t.co/qNYIudCrqb— Engineering (@XEng) November 3, 2024

Illustration: The Verge

Musk claimed that stopping people from seeing your public posts “makes no sense,” but due to a post-Musk change that stops logged-out users from scrolling even a public profile, this could make it easier for blocked users to continue harassing someone.

In October, X’s engineering account argued that people who block others could say harmful things about the blocked person who wouldn’t know about it and that the change allows for “greater transparency.” The change will still disable blocked users from following, liking, replying, reposting, DMing, or interacting with posts from the blocking party.

But as Tracy Chou, who launched the anti-harassment tool Block Party, said in a post on X, “…the point is that friction matters!! making it easy for a creeper to creep is not a good thing!!”

Another issue is whether the updated block function aligns with the rules in app stores. Apple’s guidelines say that apps with user-generated content must have “The ability to block abusive users from the service.”

Google’s guidelines are more specific:

UGC features that enable 1:1 user interaction with specific users (for example, direct messaging, tagging, mentioning, etc.) must provide an in-app functionality for blocking users.

Apps that provide access to publicly accessible UGC, such as social networking apps and blogger apps, must implement in-app functionality to report users and content, and to block users.

We’re starting to launch the block function update https://t.co/qNYIudCrqb

— Engineering (@XEng) November 3, 2024

The New York Times Tech Guild is on strike

Photo: Jakub Porzycki / NurPhoto via Getty Images

The New York Times Tech Guild went on strike at 12:01AM ET today, reports The Washington Post. The guild represents over 600 software engineers, product managers, data analysts, and designers for the media company. Its strike leaves the company bereft of crucial technical expertise just one day before Election Day, when the Times’ crucial backend systems will be under very heavy stress.
The guild has been bargaining since 2022 for its first union contract with the company. Negotiations between it and the Times hit logjams over things like a “just cause” provision that prevents the company from firing workers unless it’s for something like misconduct, as well as pay increases, pay equity, and return-to-office policies, reports the Times. The guild says on a GoFundMe page for the workers that it’s working to address racial differences in pay and discipline that disproportionately targets women and people of color.

We are on ULP strike. We gave @nytimes management months of notice of our strike deadline, we made ourselves available around the clock, but the company has decided that our members aren’t worth enough to agree to a fair contract and stop committing unfair labor practices. pic.twitter.com/jYlANW1ruw— New York Times Tech Guild (@NYTGuildTech) November 4, 2024

Times services — including games like Wordle and the NYT Cooking app — are still functional. (The guild, which will be picketing in front of the paper’s Times Square office every day, asks that readers honor its picket line by not using any of those services, according to The Washington Post.) But the Times’ election coverage is heavily tech-driven and could suffer without union worker support. A pledge signed by over 750 of its journalists said they can’t do their jobs without the workers, whose efforts are crucial when things go wrong with key coverage features like the Times election needle.
Times spokesperson Danielle Rhoades Ha told the Post last month that it has “robust plans in place to ensure that we are able to fulfill our mission and serve our readers.”

Photo: Jakub Porzycki / NurPhoto via Getty Images

The New York Times Tech Guild went on strike at 12:01AM ET today, reports The Washington Post. The guild represents over 600 software engineers, product managers, data analysts, and designers for the media company. Its strike leaves the company bereft of crucial technical expertise just one day before Election Day, when the Times’ crucial backend systems will be under very heavy stress.

The guild has been bargaining since 2022 for its first union contract with the company. Negotiations between it and the Times hit logjams over things like a “just cause” provision that prevents the company from firing workers unless it’s for something like misconduct, as well as pay increases, pay equity, and return-to-office policies, reports the Times. The guild says on a GoFundMe page for the workers that it’s working to address racial differences in pay and discipline that disproportionately targets women and people of color.

— New York Times Tech Guild (@NYTGuildTech) November 4, 2024

Times services — including games like Wordle and the NYT Cooking app — are still functional. (The guild, which will be picketing in front of the paper’s Times Square office every day, asks that readers honor its picket line by not using any of those services, according to The Washington Post.) But the Times’ election coverage is heavily tech-driven and could suffer without union worker support. A pledge signed by over 750 of its journalists said they can’t do their jobs without the workers, whose efforts are crucial when things go wrong with key coverage features like the Times election needle.

Times spokesperson Danielle Rhoades Ha told the Post last month that it has “robust plans in place to ensure that we are able to fulfill our mission and serve our readers.”

GM software boss: we have ‘high conviction’ ditching CarPlay is the right path

Photo illustration by The Verge / Photo: GM

GM’s Baris Cetinok is committed to building a custom experience to rival that of Apple and Google. Today, I’m talking with Baris Cetinok, who has one of the longest titles of any person we’ve had here on Decoder. He is, officially, General Motors’ senior vice president, software and services, product management, program management, and design.
Those are a lot of words that mean Baris is in charge of all the software in the cars that GM makes, which is a lot of cars. And if you’ve been following any of the drama in the world of car software, you know it also means Baris is the guy who has to defend GM’s decision to drop Apple CarPlay and Android Auto from most of its cars, especially EVs.
Now, Baris didn’t actually make that call, which GM announced in early 2023; he only joined GM about a year ago, after a stint running services at Apple including Apple Pay. But it’s his job to make sure that GM’s software platform, which is called Ultifi, is so good that people won’t miss CarPlay. That is a tall order. Every time we do a Decoder episode with a car person, we talk about CarPlay, and then we get an avalanche of emails from people who say they’ll never buy a car without it. See, we really do read all the emails.

Baris’ main argument for ditching these smartphone projection systems is that GM needs more control over the user experience inside of the car in order to build some of the features and services he has on the roadmap — and that handing over the displays of the car to Apple and Google simply won’t allow the company to innovate fast enough, so the company has to build its own user experience and software stack.
This is a familiar argument and a familiar approach. But it’s something we mostly hear from newer car companies — Rivian CEO RJ Scaringe said the same thing when he was on Decoder just a few months ago, but it’s very different coming from a legacy automaker like GM, which doesn’t have the luxury of starting from scratch and has to transition away from letting a huge ecosystem of suppliers write their own software for various parts of the car.
So I really wanted to know how Baris is rethinking GM’s approach to software and the teams that build it. I especially wanted to know how he plans to ship and maintain all of that software across GM’s many individual brands and models. Right now, Ultifi runs something called Google built-in, which kind of feels like having a giant Android tablet in the car. That means Baris has to deal with all the problems of giant Android tablets, like making sure they have enough memory and processing power to last for a long time, even as the demands of users and software developers increase. This is a big task.
I’ve had versions of this conversation with the CEOs of car companies before, but Baris is in charge of building this stuff, and we got really into the weeds of product and program management here. In maybe the biggest challenge of all, I asked him if anyone can actually explain all of Google’s various car platforms, and, well, you’ll hear his answer.
Okay, Baris Cetinok, GM’s SVP of software and services — and a whole bunch of other stuff. Here we go.

This transcript has been lightly edited for length and clarity.
Baris Cetinok, welcome to Decoder.
I’m delighted to be here as a longtime listener and first-time contributor. What a delight.
I’m very excited to talk to you. There’s a lot to talk about. There’s a lot going on with cars, and there’s a lot going on with GM, specifically the software you oversee. You were at Apple for a while; you actually just mentioned to me that we met at the Apple Pay launch. That’s what you worked on at Apple. You came to GM just a little over a year ago, and as of June, you have an incredible new title at GM. I’m going to read you the whole thing because I want you to explain to me what it means.
You are GM’s senior vice president of software and services, product management, program management, and design. What does that all mean?
That’s a great question. So the hypothesis around our organizational structure is that we are a functional-oriented organization. We are somewhat emulating how modern technology, especially software-driven technology firms, are structured, which is that you start with a group of people who are in the product side of the house. These are the people who think about the what: “What should we build?” And also the why: “Why should we build it?”
Their job is to be the voice of the customer inside the company. They listen to the customer, they listen to journalists, they listen to the zeitgeist, they do research themselves, read research, and develop a thesis and, eventually, a hypothesis around, “Wouldn’t it be great, wouldn’t it be cool, wouldn’t it be amazingly interesting to do X?” That’s the what and why part; that’s what product managers do.
Then, you need to think about the how, and that’s when design and engineering comes into the mix. I like the Disney term “imagineers.” These are the people who imagine and engineer things, and make those things real. You cannot ship PowerPoints, Excel spreadsheets, and Figma files. You need to turn them into real code or real beautiful designs. That’s engineering plus design.
And then program managers are the people who keep us honest, on time, on schedule, on budget, and remind us we need to make some decisions in a timely manner. They’re like the coxswain of the crew team. They’re the ones who are keeping the rhythm going, “Faster. Slow down, pay attention, more here.”
I don’t know about you, but I’m a huge sports fan. It is a team sport, and these are the top players of a basketball team’s starting five. They each have their role, they know how to defer to each other, and they know how to pass the ball. But the goal is actually just to have a winning team and build awesome products.
You’ve opened the door to an entire hour-long conversation about program managers versus product managers. I want to talk to you about a lot of other things, but we do have a lot of product managers and people who build products who listen to the show. What in your mind is the real difference between a product manager and a program manager? Because that’s not a split every company has.
Yes, and I get this question asked internally. I’ve been very fortunate, there’s been a few moments in my life where the right time, right place, and hard work met. I worked at Microsoft in the ‘90s and early 2000s.Then, I was part of the AWS group when it was just getting started at Amazon. Then, I spent a decade at Apple. One thing that’s common at all of these companies is that a strong program management culture met a strong product management culture. Product is the group that is expected to synthesize all those inputs we talked about, and put out a hypothesis. If you’re a product manager for, say, Super Cruise, you basically articulate, “Wouldn’t it be amazing if…”
It starts with simple sentences like this. It’s not some dollar sign. It’s not some ethnographic study that says, “Wouldn’t it be amazing with technology when you’re driving?” Warrior’s fan full disclosure: I drive from south of [San Francisco] to Chase Center twice a week, and 90 percent of my mileage is on [Interstate] 280, and 90-plus percent of my time is on 280. I think, “Wouldn’t it be amazing if I can be less stressed during that period? I can do other things because it can potentially be automated.” That’s the initial expression of that hypothesis.
Of course, it’s easy to say these kinds of things. It could be sci-fi, or you can make it a reality. Engineers come in, but the program managers say, “I love your vision, but now we need to divide this up into bite-sized chunks, into real and demonstrable milestones so you can truly demonstrate progress towards a real outcome and a date.” Some companies have that drum beat in a very predictable way. Every September, you must release a new phone. It sets that entire rhythm and everybody works towards that. Those are the program managers.
The same thing applies here, right? When we set out to build an amazing car like the 2025 Blazer EV, you have vehicle program managers and software program managers coming together. All these thousands of parts come together by the thousands, and build such a complex, beautiful machine called a car. That’s why you also need a strong program management culture to bring it all together, package it, and ship it. So without these four or five players playing together, it’s almost impossible to build amazing products and ship them on time.
I want to ask one more question on this. I’m trying not to get myself sidetracked into how curious I’m about this across the industry. You were at Apple. Apple is famous for not having traditional PMs, right? They have product marketing managers who are outbound to the customer. That’s very different from a bunch of other companies like Google and Meta.
How do you think about that at GM? Because you’re not really marketing the software, right? You’re marketing a car. Then GM has multiple brands with multiple expressions of the software and the car platforms. That seems like a very challenging job description.
I think you made a good observation. Apple tries to find its unicorns who are good at product management and also always has an eye for great storytelling. So I worked for a gentleman named [Greg “Joz” Joswiak] and afterwards Jennifer Bailey. They’re consummate product leaders who start with the customer value proposition, turn it into a product, and eventually ship it. So titles aside, good product managers always start with a customer backwards story. You need to have an end story in mind. Some companies like Amazon ask their product managers to write fake press releases, literally saying, “How would you like the press release to read if your product was being announced today?” So product managers actually do end up collaborating very closely with marketers because they are the stewards of the product. They know the facts and the spirit of the product.
So in a company like General Motors, we are actually very similar to some of these other companies we’re talking about. We are a hardware, software, and services company. We built beautiful industrially-designed vehicles, amazing engines — be it electrical propulsion or internal combustion — amazing interior styling, plus software and service like OnStar that’s been around for 25-plus years now, all brought into one product. When you have these types of different domains and product knowledge, it’s important to come together.
For example, I have two cars right now. I get to drive fun cars; every six months I get to drive something different. One of my cars is a Hummer EV pickup truck, and the other one could not be more on the other end of the spectrum: a not-yet released 2025 CT5-V Blackwing with it’s a six speed manual transmission. These two cars have amazingly different capabilities, with different software and capability stacks that come together. But it’s the same group of software people who are thinking about what it takes to create a performance car and also what it takes to create an amazing, hardcore off-road EV vehicle like a Hummer EV pickup truck. So, there is a way to create this type of expertise and collaborate. That’s what we’re trying to achieve here and doing it, I think, pretty well. In the last year, we had good demonstrative examples of how they came together.
First of all, well done on the Blackwing. I believe that’s the last one they’re going to make, right? I’m very jealous of this.
I actually don’t know the answer to that, but I have to say it’s an amazingly fun car.
I was going to say, if you didn’t have one Corvette in there, I was going to be mad, but a Blackwing is a reasonable substitute for a Corvette.
Actually, let’s talk about that. That’s a good framework for a bunch of questions I have here. I talked to a lot of people in the car industry, and there’s a general sense that we are undergoing what amounts to a platform change in the car industry, from ICE vehicles to EV vehicles. That creates the opportunity to rethink a bunch of experiences inside the car.
I would say this is more or less conventional wisdom. People buy an EV, they think the whole car is going to be different. This is an opportunity to actually make things different. Whereas go from one gas car to another gas car, and suddenly the whole thing’s different — that might be unexpected or strange. I’m not actually sure I buy that. I think if you buy a gas car and it has a much better infotainment system, you’re not going to be mad, right? If you buy a gas car and it has Super Cruise, I think people will be generally happy.
But there’s a bunch of stuff under that, like maybe Super Cruise is easier to deploy on an EV because you have finer-grain control over the motors, the braking system, or all the other stuff you might have mechanically in the car. Do you think that that’s true, that you need the EV transition to enable the user experience transition in the car, or are they just sort of coming along for the ride together?
First of all, I’m going to hire you as my internal evangelist. We can go on a whole tour across the company and pitch what you just pitched. But I wholeheartedly agree with you. I think what happened is that the propulsion change induced people to reimagine what cars and vehicles could do. I don’t think it was the root cause of it; it was just a spark that made people pause and start questioning what else we can reimagine in the car. However, I don’t think it’s correlation versus causation. I agree with you, propulsion change should not be the cause behind whether we should have more integrated, deeply aware experiences with our vehicles. One thing that I always try to say to people is that this is one of the most vertically integrated products you could probably buy in your life. Why wouldn’t we imagine a world where the navigation system, accelerator, HVAC, and something like Super Cruise are all talking to each other?
If you and I were just imagining a car and we knew nothing about technology, we would just describe that — that it just works, right? Forget full autonomy, you would just be like, “Hey, today I’m going to drive from here to Chase Center for the Warriors game. Help me get there.” That’s all you want. Today, we ask people to divide this into little subtasks, and you have to do all of the parts. That’s the magic of software. We take that complexity away, and we make it simpler. We automate certain tasks, and we hide certain tasks.
The thing that gets me excited about working at GM is that I get to imagine how I can tap into all the capabilities of this thing and create a deeply integrated experience. And yes, I have to say I learned it from the best. I mean, you look at a company like Apple who espouses deep vertical integration as one of the ways to create the most seamless experience. Case in point, we believe in the same thing for vehicles. and that’s why we believe we are a hardware, software, and services company rather than just a hardware company.
Apple is an interesting comparison. Obviously there was a car project there, but it appears to have not gone well. Is there a reason why GM is able to turn into a software company faster than Apple is able to turn into a car company?
One thing I came to really appreciate during my one-year tenure here is building cars is a complex system. Also, one thing that is important to appreciate is, yeah, you might get frustrated if you don’t get to post your favorite witty TikTok for an hour if there is a outage, but you don’t have that permission when you’re a safety-critical device that gets you from home to work, gets you to the hospital for your child’s first birthday, or if you need to go rush and pick up your kids from school. This needs to work every time, every day, even in snowy days and hot days reliably and safely — nonstop. That’s an amazingly high bar. There’s very few industries like that.
Healthcare is one of them, right? Imagine if you’re a Type 1 diabetic and have to use those insulin pumps that monitor your blood sugar and decide how many milligrams of insulin to give you instead of in the old days when you had to prick your finger and do your own shots. Think about those kinds of systems. These things are life-critical, safety-critical things. So one thing you come to appreciate when you start building cars is that you need to get that right. That’s non-negotiable; that is job number one. Then, you start expanding it and say, “how can I create richer experiences?” because you’re now in that vehicle.
I believe you can marry the right talent and the right players at a company. We at GM can create this kind of experience, and we have a good track record. Our cars have had software for decades now. What I think is changing is the software is now coming a bit more to the forefront in how you interact with the vehicle rather than ABS, which has been around for decades. That’s the magic of software, right? Emissions and how that’s controlled? That’s the magic of the software. How you take corners safely, that’s magical. There’s a lot of embedded software. I think what’s coming is just to the forefront and creating interactions. I believe we gathered the right group of people and the right talent to collaborate with the existing talent at GM to create those kinds of experiences. So I believe, in short, it’s possible.
When I look at the automotive industry broadly, there’s this set of startup players that have begun from an integrated platform, like Rivian and Tesla. You talk to the Rivian people and they’re proud of the fact that they built the car on ethernet, which most legacy car makers haven’t done. Then you talk to the big players like GM and Ford, and the conversation is all about reducing the number of microcontrollers across the car. We’ve got 45 ECUs across the car, we’re going to bring it down to 39, and that’s a huge victory, right? Because we’re reducing the amount of independent systems built by other suppliers.
Where is GM on this journey? I realize you’ve only been there for about a year, but it seems like you can’t get to building all of the products you want to build or all the experiences you want to build unless you dramatically re-architect the car.
This is the beauty and the hard but fun part of having such a vast and rich portfolio like we offer to GM brands. We built a Hummer as well as sedans, the Blazer, Silverado, and Bolt. We have a very large portfolio that meets the needs of very different jobs that people want to get done. What that also means is that this is a popular portfolio. It’s been around for a while. This is not a startup proof of concept with just two vehicles. That’s a little bit easier to get started since it’s easy to start from a blank sheet and say this is the new norm because there is no preceding norm. So our job is just like any successful company. And we’re not the only ones, right? Think about Windows, which probably runs on PCs that’ve been around for four decades as well as a PC you’ll buy tomorrow, and they need to power an ecosystem with thousands of permutations or millions of permutations.
We do support multiple electric architectures because that’s what we do. We do build the software for them, and we are always also looking into the future for where else we can take it. The thing is, just like any company that’s working on its next electrical or electronic architecture, you are going to have a rollout that’s going to be progressive. You cannot flip an entire portfolio with three brands and tens of different offerings on day one. Also, you don’t have to. That’s the thing. I think our customers have different needs, different expectations, and we are working through that responsibly.
My team is already actively working on the next generations of architectures because these take five years to a decade or longer. So we are on this continuum. When you get into, let’s say a 2024, 2025 Lyriq, you experience a different electric architecture then you would have if you bought a Colorado ZR2 pickup truck, and that’s okay.
By the way, customers actually don’t need to know. Our job is to hide that complexity and those choices. All you want is a reliable car that you can easily and comfortably get directions from, that routes you in such a way that you know when you need to charge, how much, where the charging stations are, that you can stream your music, answer your phone calls, etc. But let us do our job. We’ll hide the complexity, and we’ll continue to innovate and make things even better without you having to know it.
Well, the Decoder listeners want to know, so we’re going to stay right on this for one more second.
Let’s do it. Let’s unpack it.
When you talk about services broadly and what customers do or do not need to know about their cars, one of the main benefits of integrating car architectureI hear about all the time is that this is what enables over-the-air updates. The windows are malfunctioning, we’re going to ship an OTA and fix the windows. There’s a new feature we want to add to the car, so we can ship a feature and address every system in the car in one go. Drop the OTA and it’s done.
Tesla is famous for this; Rivian does this. Legacy car makers have struggled with this because you’re integrating all of these systems from all these suppliers that talk to each other, but you can’t address them directly, or your suppliers have to write the code. That’s really the thing I’m asking about here. Have you taken control of that entire architecture or is GM still using a bunch of different components from different Tier 1 suppliers?
I think the way you purposefully simplified it — I’m not saying this in a diminutive manner — is the most stark way of turning it into a binary. If you don’t have full control of everything, you don’t design your own silicon, and you don’t do this, this, or this, you cannot create a device that’s updatable. That’s actually not true. I think in the industry, we need to be a little bit cautious about actively updating cars in our current architectures. Yes, maybe some of the pieces that need to be updated are harder to upgrade, but they’re not impossible to upgrade.
This is the moment — again, I always put the onus on creators, on makers — where it’s our job to figure out the hard part. Yes, it would be amazing if I was given a blank sheet today to simply design a central compute with just one magical button. But if you were to unpack any consumer device you had in your pocket that existed for the last 15, 20 years, we all went through this type of evolution, yet we always found a way to make them updatable and kept them current. It was harder to do internally, but you as a consumer hopefully never had to feel that difference. There is great promise in moving to, for example, IEP or ethernet-based architectures, in creating more central compute units to try to get more flexibility and easier updates.
That’s on our roadmap, but I am not waiting for that day to arrive to be able to get you updates today in your Blazer EV, in your Lyriq, in your Hummer, in your Corvette. We’re doing that every day, all day. That’s why I also want to be very clear that there is not a yes-no answer for if you must wait for the future to come, and have only modern companies that do it. We are doing that every day, and we are going to continue to innovate on the electric architecture. You will see it and experience it, but one day, hopefully you don’t even notice that the silicon in your car changed because I don’t think it should matter to you.
Let me ask you the Decoder questions. Again, you’ve only been there about a year, but there has been a restructuring. GM let go of about 1,000 people. Your role has changed. You have a counterpart in David Richardson who runs engineering at GM. How is this entire group structured, and how do you account for the vastness of GM’s portfolio?
So it goes back to our opening conversation about functional orientation. So my primary partner in crime is Dave Richardson, and he oversees our engineering team as well as our IT. That’s a very important part because, like I said, they’re the imagineers. We need them and the designers to turn this into reality. One big change that happened is we’re now a primary peer group to our counterparts on the vehicle product development side. That’s actually immensely freeing because now you have the people who are imagining the vehicle because here, I use this word internally too, we’re in “service” of the vehicle. That’s what we do. Software alone is not the goal. The goal is to create an amazing vehicle. You get into it and you take that ZR1 on the track, and our president Mark Reuss took this thing up to over 230 miles per hour.
There’s so much going on in that car to make that happen, but it means all of us came together to make that happen. That’s one big structural change, that software and services are now in a primary seat at the table alongside people who are building and designing the vehicles. The other thing is we’re actively looking into how we can be leaner, more efficient, and optimized for speed because there’s this debate that takes place: How big is too big when you try to move fast? Eventually, the node connections in that network — the collaboration tax — becomes greater than the benefit of adding one more person to the nodes. That’s the thing that we always actively ask ourselves: How do we make sure this is a lean winning team that is the right size with the right players and has the right amount of connections? Not too much, not too little.
But we also want to be in a place where we say fewer yeses and make great ideas happen rather than say a lot of yeses and struggle in timing, scheduling, and other things. So a new perspective we’re bringing into the mix is one of the things that may be different. What it takes to build great software and services might be a little bit different, requiring different modus operandi than maybe some of the other parts.
You came into a structure, you changed the structure, and you have reduced the size of the structure in the interest of speed. What is the biggest difference between the structure you came into and the structure you have now?
One is just functional orientation. The second area is — and this is one of my strong beliefs — that I believe leading experts is a great model. You really want to be the experts of at least most of the things you’re overseeing. Therefore, instead of generalists managing groups, we are actively hiring or finding experts within GM to lead the expert subgroup of their domain because then there’s this really healthy debate environment, call it dialectic, seeking truth. When you’re experts at something, you have this desire to always improve and question, but when you have generalists overseeing things, they can maybe make you generally efficient, but they’re not going to be equipped to question the details.
So you have leading experts and then you have the other things we talked about, which is how do you create the trust environment where these experts trust their peer experts? They go, “You know what? I’m going to defer to you Nilay. You are better at communications and how to do better storytelling, how to make things pithier. Tell me how you think we should tell that story?” I could be dangerous and say, “Hey, I’ve been doing this for three decades, I know how to do PR.” But why should I do that? I have some of the best communications people working with me to advise me and to help me achieve that goal. I don’t think experts leading experts and experts trusting each other and passing the ball are new to GM per se, but we are iterating on those things and how we want to operate as a software and services group.
GM is a giant company. I think you report directly to CEO Mary Barra, the CEO.
Yes.
But she has to oversee how the vehicle platforms are expressed, from GMC trucks to Chevy trucks to Buick sedans, to the whole litany of GM’s cars, platforms, and brands. How do you break the tie if Cadillac wants a software feature, but Buick doesn’t?
I’ll start with, it’s hard,” but I call it “It’s hard work. Let’s do it. There’s definitely a recognition that we have a capacity. Just like any product group and engineering group, you need to be honest about the capacity you have. You need to be thoughtful about what you can do and when in terms of quality. I always say to people, “You don’t just say no. You usually say when and how.” So, when you have such a vast portfolio, you need to be commensurately sized on the software side. We won’t build an amazing vehicle that we committed to just because the software group is too busy.
However, when you’re looking at your capacity, software is not the only capacity constraint. You need to be thoughtful about suppliers. Let’s say you want to do something really magical, and you need just the right camera or just the right sensor. Well, you need to also negotiate how you get that supply delivered to the plant, which also has a capacity constraint. I have great friends who help me navigate what it takes to build a vehicle, and we’re one of the constraints but also the enablers.
So our job is to figure out if we are sized right. Do we have the right people looking at the right problems? How do we manage our capacity over multiple years — not just one year?Vehicle portfolios are a bit more like venture capital investment-style funds that need to manage or optimize your portfolio rather than a hedge fund that’s hyperactively trading. When you start imagining a vehicle like the ZR1, those ideas started probably five, seven years ago because that’s what it takes to build a great car. Sometimes what you’re doing is just a new model year of an existing car. That’s a different task. Sometimes you’re doing a midlife cycle refresh of that car.It’s definitely a complex portfolio management task, but at the end, the products you end up building are so viscerally and emotionally fun. I just call it, “It’s hard work. Let’s do it.”
How do you break the ties though?
It’s a debate club, and what happens in debate clubs stays in debate club. So, the thing is we have such amazing leaders like Mary, Mark Reuss, leaders who work for Mark, some of my brethren like Ken Morris, Josh Tavel on the design side, Michael Simcoe. I mean, these people are legends in their own right in what it takes to build a multi-car, multi-brand high quality portfolio successfully and profitably. It’s a respectful debate club. I think there’s two options:. you either agree and commit or you disagree and commit. And you get to work.
This leads me right into the big Decoder question. You have a lot of decisions to make. What is your framework to make decisions?
I think I’ve already been sprinkling in some of the clues….
I can tell you’re a listener. I can always tell when the guest is a listener. You’ve been building your way here.
Look, this is a “started at home” kind of a moment, but it is true. I grew up in a household where questions were answered by questions. It was a very Socratic method of debate and dialectic. You sought truth, and you let the debate lead you to places with your own persuasions because we all have this visceral first reaction that we think we know. It’s very important for a leader to always catch yourself and go, “I actually don’t know many things about this. Let’s do data intuition.” There’s this really famous article I love. The headline goes “Data and Intuition: Good Decisions Need Both.” That’s what it is. It is both, but you debate it because decisions are rarely so neatly binary. They’re always on a spectrum of gray, and they’re rarely deeply wrong. They’re deeply right. You debate it, you weigh it, and then, finally, it comes to a single term: conviction.
At the end, you need to make a bet and you need to have conviction. Now, this is your moment of making a bit of reality happen. It’s easy to react to empirically observable reality because then everybody can observe the same thing and come to the same conclusion. Life is easy, right? You do research and you do. The hard part is that the decisions we’re making today are crystal balling what’s going to happen in five years in terms of taste, desirability, zeitgeist, and technology. The only way you get there is to hopefully, subjectively, objectively, make good judgment calls. You make a call, you have conviction, and you put your efforts behind making that reality because there’s a lot of things that are under your control. Let’s worry about those.
There’s this term that I picked up from my days at Amazon. It’s called controllable inputs. You worry about your controllable inputs. Other things are going to happen, and you have to react. With your controllable inputs, you make your decision and you commit and start executing. I always tell people that most of success comes from relentless iterative execution rather than some magical, amazing, brilliant light bulb idea. It’s incessant dedication and doing it over and over and over again. That commitment and conviction is what gets you there with decisions.
Okay, you opened the door with conviction, so I’m going to ask you this. This is what everybody has wanted me to ask you about. The big decision, which was made before you arrived at GM, was to drop smartphone projection like CarPlay and Android Auto from GM vehicles. I’m assuming that when you took the job at GM, you agreed with this decision. You certainly haven’t undone it in your year there. Why make that decision? Why drop CarPlay and Android Auto from GM vehicles?
Because there was a belief and a hypothesis, which I genuinely believe in, that we are best positioned and owe it to our customers to create the most deeply integrated experience that you can create with the vehicle. We are not shipping devices with just monitors; we’re not a monitor company. We’re building beautifully designed, complete thoughts and complete convictions. We say, “This car is designed to do the following things awesomely.” This is Silverado, this is what it stands for and this is what it does. Let’s get to it.
When you want to create something so seamless, it’s hard to think about getting into a car and going, “Okay, so I’m doing highway trailering, but let me flip to a totally different user interface to pick my podcast. By the way, it’s a single app-obsessed interface — it’s still hard to believe. So I pick my podcast, flip back to trailering. Oh, now I can also do Super Cruise trailering. Let me manage that. Then, wait, we’re now getting into potentially Level 3, Level 4 autonomy levels that should be deeply integrated with talking to the map where the lanes lie. But wait a minute, the map that I’m using doesn’t really talk to my car.”
As a product person, you’ll never do that to yourself because it’s literally like, “Oh my God, I made my life so hard to create amazingly seamless experiences.” At some point, you need to make that bold decision and say, “I am not going to try to accommodate and figure out how to make all of these work. I’m going to just burn the bridges and burn the ships and commit.” We are going to create a deeply vertical, harmonious experience that works across the vehicle that is optimized for my vehicle.
I can appreciate when you’re in that one-off rental car that you don’t want to be driving but is the only car that’s available when you land, be my guest. Use projection. But the car that I researched for months and decided to buy for its interior, exteriors, its propulsion, and every feature that it does for me? We bought a Colorado ZR2 Bison. We love cars in our family; as you can see, we’re very outdoorsy. We go gravel biking and rock climbing in Yosemite with my son. I want that car to be true and true optimized for what I bought it for, not some lowest common denominator moment. That’s why it was not an easy decision to embrace when I joined, I’ll be very honest with you. At first, it’s a seemingly hard decision and maybe unpopular decision, but it is the right decision because we are here building an end-to-end integrated product for you.
I just have to say this: a lot of times when I land at the airport and I get in the rental car, it’s a white Chevy Malibu. Are you going to leave smartphone projection alone in the fleet cars that get put in rental lots?
I’ll take the suggestion back to our senior leadership team. Let’s see what happens.
I’m just curious because you’re describing one very specific thing that people like smartphone projection for.
Yeah, you’re right.
It is a big market for GM. That’s the kind of fine-grained debate about smartphone projection I get from Verge readers and Decoder listeners. There are all these places where that idea sounds great, and there’s this whole list of other places where that sounds horrible actually. When I borrow my parents’ car, I do not want to live in their Google account. I just want to send my Spotify to the screen.
It seems like you’re burning the bridges, but you’re burning the bridges to a lot of other experiences that people have in cars.
The one thing that I believe and encourage our partners — and we work very closely with Apple, Google, Amazon, Spotify, and others — to believe is that some of those pieces of information and preferences should be accessible to other devices that you choose to use. I don’t think they need to be locked in this walled garden of a single domain. I personally believe I should be able to take my podcast preferences from one podcast app to another. I always find it interesting that some of these very generic preferences I declare are locked and loaded, especially for non-specific content.
I appreciate that a certain show is only on Netflix, let’s say — respect — but most song music catalogs, podcasts, mapping information, your home address, your work address, these are pretty generic pieces of information that you probably want to just enable in any vehicle you walk into to make that experience more personal. So that’s one thing that our team is also working on. How can we actually get that information to power any bespoke experience that’s in a vehicle rather than having to relinquish the full vertical experience to a totally different one?
So I just spent a weekend test-driving a Blazer EV. A lot of fun. That whole stack is built on Android, right? It’s Google services, the Play Store, apps. It wanted me to log into my Google account. That’s where all of that information came from. Then, it wanted me to log into my Spotify account, which is where a bunch of that information came from. At some point, you’ve kind of just put an Android tablet in the middle of my car. Why build on Android? What you’re describing is a big vision. and then the way it’s expressed right now is, “Well, here’s Android.”
The way I look at it is if we were to do a startup today, we would probably start playing with existing building blocks, right? If you and I were to do, let’s say, an overlanding app, you probably wouldn’t come to me and say, “I have an idea. First we should get a data center and get a bunch of racks.” We would probably go to AWS, and then you’d say, “Hey, for developing this app, we’re going to use React. There’s some existing open-source libraries, let’s take those.” And let’s say we were also in the business of using some kind of map. We would be like, “Well, when we’re working on this platform, we’ll use this map versus here where we don’t need to actually integrate a third-party map.” We’re not going to build our own map. Think about it that way.
As a software group and software developing group, we don’t have to build every single building block. You have capabilities out there that are already given to you. Android is a building block; it’s an open source system. In addition to that, Google provides building blocks that you can tap into and build around. We’re working with other companies and seeing if they’ll be interested in building their building blocks and their apps into this environment where we get to create the experience.
So, I look at this as just enabling capabilities and wanting these moments you talked about,how you need to log in here to get the information from that ecosystem and make it more personal. I’m going to challenge my team. I think we should make that super magical and simple. Why can’t we just detect that you have your phone in your pocket and do a very simple handshake instead of scanning a QR code? Today that’s the most seamless way of doing it, but there’s many different ways to do it.
The same thing with if you log in once, what if you log into one plus another GM built car? Can we carry that over? Can we create one authentication system around GM ID that carries some of your other credentials? All of these are, in my opinion, solvable problems, and we’re now setting out to do that. It takes dedication and having this platform thinking to be able to, one at a time and with dedication, remove all of these moments of friction you articulated. I agree with you, and we’re going to make it simpler and simpler.
One thing I’ll say though is, thankfully, most people don’t change their cars as often as you do. You’re changing your cars as often you probably change your shirts. Most people buy a car and love it for three, four years. I used to say this in other places I work, most people don’t upgrade the operating system of their phones every week with a new build. Most of us hold some of our devices for longer. Maybe you wish you didn’t have to do the setup routine, but you do it once and now it’s your car, it’s your vehicle, it’s your data, it’s your choices.
I want to come back to this timeline of three or four years. But first, I want to ask you a very challenging question — I will warn you, not every car executive who’s been on Decoder has been able to answer.
I’m listening.
So this is the challenge: Can you name all of Google’s car platforms and describe what they do?
Let me give you the phone number of a Google executive.
I don’t think they can do it either. I want to be very clear about this.
Look, I’ll say this, we have a great, tight partnership with them, and what I try to do when you look at our–
No, no, no, you’ve got to try.
I’m going to try.
Which ones do you use?
We definitely use Google built-in and Google Automotive Services. And this is the thing: most people don’t know macOS is actually built on a version of Unix. So when you actually start splitting the layers of the cakes, there is definitely open source or Linux, and Android is actually a Linux derivative, when you think about it. Let’s get all nerdy here. When you start looking at it, all of this is actually based on some other thing.
So we definitely use Google Built-In, and we use Google Automotive Services to create the experiences, but we create our own experience around it. With the Blazer EV, there are these moments to make that more personal. You can use Google Maps with all the things that you’ve done in Google Maps before. If you would like, you can sign in with your Google account so your homework or the last search you did on your phone also just appears in your recents. So those are the kind of moments where we hook into Google services. But yeah, it’s hard to build popular building blocks like Google does or AWS does and keep all their branding and naming right. So I hear you.
What’s the difference between Google Built-In and Android Automotive? Which is the Android that runs–
Android Automotive Services.
Well, there’s Google Automotive Services, which is basically the Play Store and the apps, and then there’s Google Built-In, which is what you have.
Correct.
And then there’s Android Automotive? I’ve been trying to keep this correct and clear in my head. I feel like you would know.
No, I do know.
You’re chipping the software–
I do know, I do know. What I’m laughing about is that consumers don’t need to know these things–
But I need to know. I’m dying to know. This is why you’re here.
So I’ll give you an answer that is going to be less satisfying than you want. We have different model years and different brands, some of them have been with us longer and some of them are brand new vehicles. You get into a Lyriq, you get a different experience. So our goal, first of all, is to unify those experiences. That’s why I’m here, that’s why Dave is here, that’s why all of our teams are here: to actually take away that complexity that you’re describing. You can actually see where in which vehicles we’re using different combinations of services, but again, I think it’s pretty opaque from a consumer perspective. I don’t think they need to know these differences, but I’ll point you to a great white paper that tries to explain it all.
Oh, I’ve read the white papers, I’ve seen the blog. I guarantee you more people are going to send me more blog posts. I’ll just keep asking. One day we’re going to figure this out.
What does the deal with Google to use its stuff look like? On a smartphone, that is the subject of pretty ferocious regulatory interests, litigation, you have to take Play services, you’ve got to use its app, and they get a 30% split. Is it the same on the car?
The commercial details of it are definitely confidential. I don’t think I should be in the business of disclosing our partnership details. But from our perspective — just like you articulated — it’s complex in that there are different types of experiences and building blocks, and depending on how you use those different blocks, there are different terms and conditions of how you integrate, operate, and update them.
So there is not a simple neat answer because this is unlike Windows. This is not a standard Windows PC where you buy a Windows license and you’re off to the races as an OEM. The cars we create are far more complex and capable, so therefore, a good portion of the cars are using things that are not directly licensed from Google while some portions are directly licensed for Google — where Google Maps is Google Maps or Google Assistant is Google Assistant. Every one of those permutations as they come together is different and unique.
The reason I ask that question is because one of the reasons so many car makers are interested in taking control away from smartphone projection and building their own user interfaces is what you have described: to offer more services, offer more subscription services in the car, offer more transactions on the screen, taking 30 percent of every transaction that happens on screen in the car, offering streaming services.
I’ve heard every version of these ideas, and if your economics include having to pay Google a fee for Android apps operating in your car on top of the fees you are extracting, it feels like there’s a real tension there, with Google as an operating system vendor or a services vendor wanting a fee or a percentage split on top of the one the GM wants from an app developer. And then the economics for an app developer to put an app on your car platform starts to fall apart pretty fast.
We see this play out in smartphones, right? I think the day before we were speaking, Hulu said, “we’re not going to allow sign-ups on iOS directly anymore because we don’t want to pay Apple a 30 percent split.” I see this tension on smartphones with just one intermediary in Apple and Google. In your car there’s two, right? There’s Apple, here’s Google, and then there’s potentially GM.
So let me parse this question a little bit. I’ll start with that our primary motivation in every conversation we have is to first create a great customer experience. I have this simple axiom: “great products also usually make great businesses.” But you always start with a great product. In everything we discuss, we’re always trying to continuously optimize how we create a great experience for you. That’s it, that’s the primary goal.
The second thing is the type of services we are building that I am really excited about. The good thing is that we are a company that has already proven that model with OnStar. We have millions of subscribers who are using it to extend their vehicle experience in a way that is expected from the vehicle, which is how do you make this safer for me? How do you help me recover my car if there was ever a theft? How do we help you insure your car easier because that is, again, integral to your car. Super Cruise is a great example, right? Super Cruise is ever improving. We continuously add more and more roads, for example, which means we’re adding a concrete value. We just upped that to 750,000 miles across the U.S. That means we’re investing, therefore there’s value added and value created, so consumers are willing to pay for that subscription.
Those are the types of services I’m deeply interested in: things that are extending my experience as a vehicle. I think that’s ours, that’s ours to build and ours to develop. Then there are some opportunities like you articulated around entertainment, around gaming, etc. But to me they come next. I really, and we are obsessed with extending the vehicle experiences first because we know how to do that really well and there are some domains.
I’m not here to compete with Spotify or Apple Music. They do a fantastic job servicing all of our music listening needs. We work with them, we bring their apps to our platform, and we run it. So I am not too concerned about that part that you articulated: “How do you create value? How do you capture value? How do you split the economics?” I think those are all solvable problems, but it all starts with the experiences I owe you as a consumer who buys a beautiful Escalade IQ in a month or two. What are the experiences I should create? When you get into that car, you’re going to see this pillar-to-pillar display with front control and back control. We have so many ideas of what experiences we can create for you there before we get into this area of revenue and shared debates around a gaming app.
Earlier, you mentioned three to four years, and the reason that immediately lit up my brain is because when listeners ask me about CarPlay, a thing they say to me is, “Well, if you’re going to put a giant Android tablet in my car, that means I have a giant Android tablet in my car and Android tablets get slow. Three years from now, that will be slow, and with smartphone projection, it’s always running off my phone, which is always fast.”
How do you deal with, “Okay, we’ve put all of the user interface on an SoC in a car that might have to last for a decade?” Do you just over-spec them to make sure they have enough headroom to last that long? Do you just assume people are going to turn them in on three-year leases? What’s the story?
You deal with multiple things. First of all, I’m a huge believer in creating elegant software. We’re not just simply running some off-the-shelf Android and just conceding the entire thing in a non-optimized way. We’re directly involved. We actually are using that as a building block and we’re creating our experience. Therefore, the onus is on us. How do we create more optimal and optimized code? I’m a huge believer that you can continuously improve your code from the day you ship till the day you update it. We are on a continuous optimization journey. This thing doesn’t stay stable and just collects on top of that additional debt.
In the other areas, you’re absolutely right. You do look at the compute investment you’re making and you want to give yourself headroom because today we’re doing Level 2 [autonomy], which is, as I call it, “eyes-on, hands-off driving” with Super Cruise. But we are also the company that has already invested in a fully autonomous future, which is with Level 4. There is a zone of Level 3 autonomy. Well, if you want to live on that journey, you want to create a system that has some headroom or is extensible and expandable as you add new capabilities. Maybe you don’t want to over-provision from day one because you’re not going to take advantage of it.
I never worked in this industry, but I am familiar with it because I was a gamer — I have less time to do gaming now — but the same thing applied when you were creating the next PS5 versus the PS6. You knew this platform was going to be there for many years to come, and you were trying to create just the right balance of provision with room for extendability, optimized software, and bringing content in that feels fresh, new, and always improving. That’s exactly what we’re doing with our vehicles. If you bought a Lyriq last year, the Lyriqyou’re driving today has newer software, is more optimized, and probably running faster than when you bought it because we’re continuously improving it.
You mentioned other platforms. I always think about Windows Phone for some reason when people talk about platforms like this. Again, when listeners tell me why they like CarPlay, which they’re constantly telling me about why they like CarPlay, one of the things they say is, “All of the apps on my phone support CarPlay.” I have one listener who told me he routinely switches between six and seven apps in CarPlay while he’s commuting. It’s like podcast apps, ebook apps, all this stuff, and he’s just flipping. I find that incredible. I use two apps, but so it goes.
That means you need the app support. If you want to sell that listener a Blazer or a Lyriq, he’s got to sit down, open the Play Store — which is not great for discovery right now, I’ll tell you — and he’s got to search for all of his seven apps and he’s got to find them all, which is the problem Windows Phone had, right? The volume of apps wasn’t there, and the library wasn’t there. Have you thought about solving that problem? Is that GM’s problem? Is that Google’s problem? How are you addressing that?
I think first of all, anything that happens in our cars is our problem. That’s how I look at it. There’s that finding that happy medium between what are the most popular apps we believe should be there to complete your experience, and sometimes those ones — and this even applies today — that don’t even have apps. Even with your phone you sometimes need to go punch through to the internet and go to the browser and do it, right? These types of different domains exist in our lives in other places too. We are working actively with the people we believe provide the core apps, maybe the top 10. It’s even more stark when you start talking to people. What are the must-have apps while you are driving or while you are a passenger when you’re riding in our vehicles? That list is actually far shorter than the long tail of things we do.
I think you’re touching up on an interesting area that I am also interested in, which is if there is a way to bring vehicle-optimized versions of certain apps. Let’s say you’re a surfer, there’s this great app called Surfline. Wouldn’t it be fascinating if you had a vehicle-optimized version of the Surfline app that gave you information about what the tide conditions are, wind, etc. but in a smart way? Now, let’s go a little bit into the future. I believe this rabbit hole of vertical, micro-segmented apps is going to go under a change too, especially with voice interfaces becoming far more intuitive than ever before because of large language models.
For me, I think the biggest change with large language models is that I can finally be understood by the machine despite the fact I did not phrase myself perfectly. That device is going to be able to interpret my intent and give me answers. The reason I think we all default to a more touch user interface today is that we as users have to do a lot of the disambiguation instead of the machine understanding what you want. You are going through things, when all you want to know actually is this. You don’t need an app. If you want to drive to Aptos, California, to surf in an hour: “Tell me the surf conditions.” It’s a two-sentence answer. Do I really need the app for that?
Apple’s solution to this is next-generation CarPlay where it will send more information to more screens, where it will help car makers design their center console, help car makers design the instrument cluster and other screens. Your old boss, [Greg Joswiak], is a big advocate of this. It’s been taken up a little bit, but not a lot. Have you looked at it? Would you use next-gen CarPlay?
I look at almost everything. I’ve driven almost every competitor’s car under the sun, and I have all the devices that you probably also carry in your pocket to learn, test, and understand. We already made our decision about the experience we want to create, which is going to be a deeply vertically-integrated experience that GM designs, builds, maintains, updates, and innovates on for our customers. That’s our direction.
There’s a really interesting A/B test happening in the market right now. You’ve got the Blazer EV, which has your approach. There’s the Honda Prologue, which is substantially the same platform, and the Acura ZDX, which is substantially the same EV platform. They support CarPlay. The Prologue is kind of a hit, I think a surprise hit for Honda. Are you keeping track of that? Do you think that’s attributable to CarPlay? Is that difference in branding?
Like I said, we’re looking at the competition, but we’re probably a customer and quality-obsessed company rather than competition-obsessed company. We’re definitely a highly competitive company, but it goes back to something we talked about: It’s about conviction. We have high conviction that this is the right path for us to take to create truly interior experiences for our customers. Like I said, an easier decision would be to ask, “Why make that effort?”
But we have a strong conviction that effort pays off in a better customer experience. You get the most out of your vehicle because now we’re the company that builds the vehicle and is also creating the infotainment experience, the cluster experience, the app, and everything. We’re going to build that one day and maybe a voice assistant on top of it. The only way you can create that end-to-end magic is to have a strong conviction that you want to own all of these. It’s hard to create those seamless experiences. You’re going to always feel the seams, and I don’t think you should be in the business of feeling the seams when you buy a $30,000 to $100,000 car. It’s our job to make it beautiful and seamless.
We have a lot of listeners who say, “Look, I’m just not going to buy a car without CarPlay, and that’s fine.” Point them to something that will make them change their mind, that will make them reconsider. What’s something coming up where you’re going to say, “I can build you an experience that will make you reconsider?”
I don’t think you need to wait for the future. I think Super Cruise and how it integrates natively with the–
You can get Super Cruise on a [Cadillac] Lyriq with CarPlay today.
Well, the thing is that when you are using the native integration with the map that knows where you’re going to be, what are the road conditions, what’s happening, where are the construction sites, and it ties it back to the battery system and says, “By the time you get to your destination and return, your battery’s going to be at this level. Here is the energy efficient route you can take.”
I cannot create those magical little dots connecting to each other if I had to go back and forth between a navigation app running on another platform, and that’s the thing: you don’t need to wait for the future to create that. The other things that we want to create are similar things, especially when we start introducing Level 3 autonomy, hands off, eyes off, and even with robotaxis. That type of integration is a prerequisite, not an optional feature.
Well, Baris, thank you so much for coming on Decoder. You got to come back so we can do a full hour on robotaxis, and a full hour of the difference between product and program management.
Nilay, I really enjoyed our conversation. Great questions, and I’m going to listen to this episode too. I hope it turns out pretty great.
We can all learn something. Maybe we’ll get that blog post from Google. I’ll let you know. Thank you, Baris.
Thanks so much. Take care.

Photo illustration by The Verge / Photo: GM

GM’s Baris Cetinok is committed to building a custom experience to rival that of Apple and Google.

Today, I’m talking with Baris Cetinok, who has one of the longest titles of any person we’ve had here on Decoder. He is, officially, General Motors’ senior vice president, software and services, product management, program management, and design.

Those are a lot of words that mean Baris is in charge of all the software in the cars that GM makes, which is a lot of cars. And if you’ve been following any of the drama in the world of car software, you know it also means Baris is the guy who has to defend GM’s decision to drop Apple CarPlay and Android Auto from most of its cars, especially EVs.

Now, Baris didn’t actually make that call, which GM announced in early 2023; he only joined GM about a year ago, after a stint running services at Apple including Apple Pay. But it’s his job to make sure that GM’s software platform, which is called Ultifi, is so good that people won’t miss CarPlay. That is a tall order. Every time we do a Decoder episode with a car person, we talk about CarPlay, and then we get an avalanche of emails from people who say they’ll never buy a car without it. See, we really do read all the emails.

This is a familiar argument and a familiar approach. But it’s something we mostly hear from newer car companies — Rivian CEO RJ Scaringe said the same thing when he was on Decoder just a few months ago, but it’s very different coming from a legacy automaker like GM, which doesn’t have the luxury of starting from scratch and has to transition away from letting a huge ecosystem of suppliers write their own software for various parts of the car.

So I really wanted to know how Baris is rethinking GM’s approach to software and the teams that build it. I especially wanted to know how he plans to ship and maintain all of that software across GM’s many individual brands and models. Right now, Ultifi runs something called Google built-in, which kind of feels like having a giant Android tablet in the car. That means Baris has to deal with all the problems of giant Android tablets, like making sure they have enough memory and processing power to last for a long time, even as the demands of users and software developers increase. This is a big task.

I’ve had versions of this conversation with the CEOs of car companies before, but Baris is in charge of building this stuff, and we got really into the weeds of product and program management here. In maybe the biggest challenge of all, I asked him if anyone can actually explain all of Google’s various car platforms, and, well, you’ll hear his answer.

Okay, Baris Cetinok, GM’s SVP of software and services — and a whole bunch of other stuff. Here we go.

This transcript has been lightly edited for length and clarity.

Baris Cetinok, welcome to Decoder.

I’m delighted to be here as a longtime listener and first-time contributor. What a delight.

I’m very excited to talk to you. There’s a lot to talk about. There’s a lot going on with cars, and there’s a lot going on with GM, specifically the software you oversee. You were at Apple for a while; you actually just mentioned to me that we met at the Apple Pay launch. That’s what you worked on at Apple. You came to GM just a little over a year ago, and as of June, you have an incredible new title at GM. I’m going to read you the whole thing because I want you to explain to me what it means.

You are GM’s senior vice president of software and services, product management, program management, and design. What does that all mean?

That’s a great question. So the hypothesis around our organizational structure is that we are a functional-oriented organization. We are somewhat emulating how modern technology, especially software-driven technology firms, are structured, which is that you start with a group of people who are in the product side of the house. These are the people who think about the what: “What should we build?” And also the why: “Why should we build it?”

Their job is to be the voice of the customer inside the company. They listen to the customer, they listen to journalists, they listen to the zeitgeist, they do research themselves, read research, and develop a thesis and, eventually, a hypothesis around, “Wouldn’t it be great, wouldn’t it be cool, wouldn’t it be amazingly interesting to do X?” That’s the what and why part; that’s what product managers do.

Then, you need to think about the how, and that’s when design and engineering comes into the mix. I like the Disney term “imagineers.” These are the people who imagine and engineer things, and make those things real. You cannot ship PowerPoints, Excel spreadsheets, and Figma files. You need to turn them into real code or real beautiful designs. That’s engineering plus design.

And then program managers are the people who keep us honest, on time, on schedule, on budget, and remind us we need to make some decisions in a timely manner. They’re like the coxswain of the crew team. They’re the ones who are keeping the rhythm going, “Faster. Slow down, pay attention, more here.”

I don’t know about you, but I’m a huge sports fan. It is a team sport, and these are the top players of a basketball team’s starting five. They each have their role, they know how to defer to each other, and they know how to pass the ball. But the goal is actually just to have a winning team and build awesome products.

You’ve opened the door to an entire hour-long conversation about program managers versus product managers. I want to talk to you about a lot of other things, but we do have a lot of product managers and people who build products who listen to the show. What in your mind is the real difference between a product manager and a program manager? Because that’s not a split every company has.

Yes, and I get this question asked internally. I’ve been very fortunate, there’s been a few moments in my life where the right time, right place, and hard work met. I worked at Microsoft in the ‘90s and early 2000s.Then, I was part of the AWS group when it was just getting started at Amazon. Then, I spent a decade at Apple. One thing that’s common at all of these companies is that a strong program management culture met a strong product management culture. Product is the group that is expected to synthesize all those inputs we talked about, and put out a hypothesis. If you’re a product manager for, say, Super Cruise, you basically articulate, “Wouldn’t it be amazing if…”

It starts with simple sentences like this. It’s not some dollar sign. It’s not some ethnographic study that says, “Wouldn’t it be amazing with technology when you’re driving?” Warrior’s fan full disclosure: I drive from south of [San Francisco] to Chase Center twice a week, and 90 percent of my mileage is on [Interstate] 280, and 90-plus percent of my time is on 280. I think, “Wouldn’t it be amazing if I can be less stressed during that period? I can do other things because it can potentially be automated.” That’s the initial expression of that hypothesis.

Of course, it’s easy to say these kinds of things. It could be sci-fi, or you can make it a reality. Engineers come in, but the program managers say, “I love your vision, but now we need to divide this up into bite-sized chunks, into real and demonstrable milestones so you can truly demonstrate progress towards a real outcome and a date.” Some companies have that drum beat in a very predictable way. Every September, you must release a new phone. It sets that entire rhythm and everybody works towards that. Those are the program managers.

The same thing applies here, right? When we set out to build an amazing car like the 2025 Blazer EV, you have vehicle program managers and software program managers coming together. All these thousands of parts come together by the thousands, and build such a complex, beautiful machine called a car. That’s why you also need a strong program management culture to bring it all together, package it, and ship it. So without these four or five players playing together, it’s almost impossible to build amazing products and ship them on time.

I want to ask one more question on this. I’m trying not to get myself sidetracked into how curious I’m about this across the industry. You were at Apple. Apple is famous for not having traditional PMs, right? They have product marketing managers who are outbound to the customer. That’s very different from a bunch of other companies like Google and Meta.

How do you think about that at GM? Because you’re not really marketing the software, right? You’re marketing a car. Then GM has multiple brands with multiple expressions of the software and the car platforms. That seems like a very challenging job description.

I think you made a good observation. Apple tries to find its unicorns who are good at product management and also always has an eye for great storytelling. So I worked for a gentleman named [Greg “Joz” Joswiak] and afterwards Jennifer Bailey. They’re consummate product leaders who start with the customer value proposition, turn it into a product, and eventually ship it. So titles aside, good product managers always start with a customer backwards story. You need to have an end story in mind. Some companies like Amazon ask their product managers to write fake press releases, literally saying, “How would you like the press release to read if your product was being announced today?” So product managers actually do end up collaborating very closely with marketers because they are the stewards of the product. They know the facts and the spirit of the product.

So in a company like General Motors, we are actually very similar to some of these other companies we’re talking about. We are a hardware, software, and services company. We built beautiful industrially-designed vehicles, amazing engines — be it electrical propulsion or internal combustion — amazing interior styling, plus software and service like OnStar that’s been around for 25-plus years now, all brought into one product. When you have these types of different domains and product knowledge, it’s important to come together.

For example, I have two cars right now. I get to drive fun cars; every six months I get to drive something different. One of my cars is a Hummer EV pickup truck, and the other one could not be more on the other end of the spectrum: a not-yet released 2025 CT5-V Blackwing with it’s a six speed manual transmission. These two cars have amazingly different capabilities, with different software and capability stacks that come together. But it’s the same group of software people who are thinking about what it takes to create a performance car and also what it takes to create an amazing, hardcore off-road EV vehicle like a Hummer EV pickup truck. So, there is a way to create this type of expertise and collaborate. That’s what we’re trying to achieve here and doing it, I think, pretty well. In the last year, we had good demonstrative examples of how they came together.

First of all, well done on the Blackwing. I believe that’s the last one they’re going to make, right? I’m very jealous of this.

I actually don’t know the answer to that, but I have to say it’s an amazingly fun car.

I was going to say, if you didn’t have one Corvette in there, I was going to be mad, but a Blackwing is a reasonable substitute for a Corvette.

Actually, let’s talk about that. That’s a good framework for a bunch of questions I have here. I talked to a lot of people in the car industry, and there’s a general sense that we are undergoing what amounts to a platform change in the car industry, from ICE vehicles to EV vehicles. That creates the opportunity to rethink a bunch of experiences inside the car.

I would say this is more or less conventional wisdom. People buy an EV, they think the whole car is going to be different. This is an opportunity to actually make things different. Whereas go from one gas car to another gas car, and suddenly the whole thing’s different — that might be unexpected or strange. I’m not actually sure I buy that. I think if you buy a gas car and it has a much better infotainment system, you’re not going to be mad, right? If you buy a gas car and it has Super Cruise, I think people will be generally happy.

But there’s a bunch of stuff under that, like maybe Super Cruise is easier to deploy on an EV because you have finer-grain control over the motors, the braking system, or all the other stuff you might have mechanically in the car. Do you think that that’s true, that you need the EV transition to enable the user experience transition in the car, or are they just sort of coming along for the ride together?

First of all, I’m going to hire you as my internal evangelist. We can go on a whole tour across the company and pitch what you just pitched. But I wholeheartedly agree with you. I think what happened is that the propulsion change induced people to reimagine what cars and vehicles could do. I don’t think it was the root cause of it; it was just a spark that made people pause and start questioning what else we can reimagine in the car. However, I don’t think it’s correlation versus causation. I agree with you, propulsion change should not be the cause behind whether we should have more integrated, deeply aware experiences with our vehicles. One thing that I always try to say to people is that this is one of the most vertically integrated products you could probably buy in your life. Why wouldn’t we imagine a world where the navigation system, accelerator, HVAC, and something like Super Cruise are all talking to each other?

If you and I were just imagining a car and we knew nothing about technology, we would just describe that — that it just works, right? Forget full autonomy, you would just be like, “Hey, today I’m going to drive from here to Chase Center for the Warriors game. Help me get there.” That’s all you want. Today, we ask people to divide this into little subtasks, and you have to do all of the parts. That’s the magic of software. We take that complexity away, and we make it simpler. We automate certain tasks, and we hide certain tasks.

The thing that gets me excited about working at GM is that I get to imagine how I can tap into all the capabilities of this thing and create a deeply integrated experience. And yes, I have to say I learned it from the best. I mean, you look at a company like Apple who espouses deep vertical integration as one of the ways to create the most seamless experience. Case in point, we believe in the same thing for vehicles. and that’s why we believe we are a hardware, software, and services company rather than just a hardware company.

Apple is an interesting comparison. Obviously there was a car project there, but it appears to have not gone well. Is there a reason why GM is able to turn into a software company faster than Apple is able to turn into a car company?

One thing I came to really appreciate during my one-year tenure here is building cars is a complex system. Also, one thing that is important to appreciate is, yeah, you might get frustrated if you don’t get to post your favorite witty TikTok for an hour if there is a outage, but you don’t have that permission when you’re a safety-critical device that gets you from home to work, gets you to the hospital for your child’s first birthday, or if you need to go rush and pick up your kids from school. This needs to work every time, every day, even in snowy days and hot days reliably and safely — nonstop. That’s an amazingly high bar. There’s very few industries like that.

Healthcare is one of them, right? Imagine if you’re a Type 1 diabetic and have to use those insulin pumps that monitor your blood sugar and decide how many milligrams of insulin to give you instead of in the old days when you had to prick your finger and do your own shots. Think about those kinds of systems. These things are life-critical, safety-critical things. So one thing you come to appreciate when you start building cars is that you need to get that right. That’s non-negotiable; that is job number one. Then, you start expanding it and say, “how can I create richer experiences?” because you’re now in that vehicle.

I believe you can marry the right talent and the right players at a company. We at GM can create this kind of experience, and we have a good track record. Our cars have had software for decades now. What I think is changing is the software is now coming a bit more to the forefront in how you interact with the vehicle rather than ABS, which has been around for decades. That’s the magic of software, right? Emissions and how that’s controlled? That’s the magic of the software. How you take corners safely, that’s magical. There’s a lot of embedded software. I think what’s coming is just to the forefront and creating interactions. I believe we gathered the right group of people and the right talent to collaborate with the existing talent at GM to create those kinds of experiences. So I believe, in short, it’s possible.

When I look at the automotive industry broadly, there’s this set of startup players that have begun from an integrated platform, like Rivian and Tesla. You talk to the Rivian people and they’re proud of the fact that they built the car on ethernet, which most legacy car makers haven’t done. Then you talk to the big players like GM and Ford, and the conversation is all about reducing the number of microcontrollers across the car. We’ve got 45 ECUs across the car, we’re going to bring it down to 39, and that’s a huge victory, right? Because we’re reducing the amount of independent systems built by other suppliers.

Where is GM on this journey? I realize you’ve only been there for about a year, but it seems like you can’t get to building all of the products you want to build or all the experiences you want to build unless you dramatically re-architect the car.

This is the beauty and the hard but fun part of having such a vast and rich portfolio like we offer to GM brands. We built a Hummer as well as sedans, the Blazer, Silverado, and Bolt. We have a very large portfolio that meets the needs of very different jobs that people want to get done. What that also means is that this is a popular portfolio. It’s been around for a while. This is not a startup proof of concept with just two vehicles. That’s a little bit easier to get started since it’s easy to start from a blank sheet and say this is the new norm because there is no preceding norm. So our job is just like any successful company. And we’re not the only ones, right? Think about Windows, which probably runs on PCs that’ve been around for four decades as well as a PC you’ll buy tomorrow, and they need to power an ecosystem with thousands of permutations or millions of permutations.

We do support multiple electric architectures because that’s what we do. We do build the software for them, and we are always also looking into the future for where else we can take it. The thing is, just like any company that’s working on its next electrical or electronic architecture, you are going to have a rollout that’s going to be progressive. You cannot flip an entire portfolio with three brands and tens of different offerings on day one. Also, you don’t have to. That’s the thing. I think our customers have different needs, different expectations, and we are working through that responsibly.

My team is already actively working on the next generations of architectures because these take five years to a decade or longer. So we are on this continuum. When you get into, let’s say a 2024, 2025 Lyriq, you experience a different electric architecture then you would have if you bought a Colorado ZR2 pickup truck, and that’s okay.

By the way, customers actually don’t need to know. Our job is to hide that complexity and those choices. All you want is a reliable car that you can easily and comfortably get directions from, that routes you in such a way that you know when you need to charge, how much, where the charging stations are, that you can stream your music, answer your phone calls, etc. But let us do our job. We’ll hide the complexity, and we’ll continue to innovate and make things even better without you having to know it.

Well, the Decoder listeners want to know, so we’re going to stay right on this for one more second.

Let’s do it. Let’s unpack it.

When you talk about services broadly and what customers do or do not need to know about their cars, one of the main benefits of integrating car architectureI hear about all the time is that this is what enables over-the-air updates. The windows are malfunctioning, we’re going to ship an OTA and fix the windows. There’s a new feature we want to add to the car, so we can ship a feature and address every system in the car in one go. Drop the OTA and it’s done.

Tesla is famous for this; Rivian does this. Legacy car makers have struggled with this because you’re integrating all of these systems from all these suppliers that talk to each other, but you can’t address them directly, or your suppliers have to write the code. That’s really the thing I’m asking about here. Have you taken control of that entire architecture or is GM still using a bunch of different components from different Tier 1 suppliers?

I think the way you purposefully simplified it — I’m not saying this in a diminutive manner — is the most stark way of turning it into a binary. If you don’t have full control of everything, you don’t design your own silicon, and you don’t do this, this, or this, you cannot create a device that’s updatable. That’s actually not true. I think in the industry, we need to be a little bit cautious about actively updating cars in our current architectures. Yes, maybe some of the pieces that need to be updated are harder to upgrade, but they’re not impossible to upgrade.

This is the moment — again, I always put the onus on creators, on makers — where it’s our job to figure out the hard part. Yes, it would be amazing if I was given a blank sheet today to simply design a central compute with just one magical button. But if you were to unpack any consumer device you had in your pocket that existed for the last 15, 20 years, we all went through this type of evolution, yet we always found a way to make them updatable and kept them current. It was harder to do internally, but you as a consumer hopefully never had to feel that difference. There is great promise in moving to, for example, IEP or ethernet-based architectures, in creating more central compute units to try to get more flexibility and easier updates.

That’s on our roadmap, but I am not waiting for that day to arrive to be able to get you updates today in your Blazer EV, in your Lyriq, in your Hummer, in your Corvette. We’re doing that every day, all day. That’s why I also want to be very clear that there is not a yes-no answer for if you must wait for the future to come, and have only modern companies that do it. We are doing that every day, and we are going to continue to innovate on the electric architecture. You will see it and experience it, but one day, hopefully you don’t even notice that the silicon in your car changed because I don’t think it should matter to you.

Let me ask you the Decoder questions. Again, you’ve only been there about a year, but there has been a restructuring. GM let go of about 1,000 people. Your role has changed. You have a counterpart in David Richardson who runs engineering at GM. How is this entire group structured, and how do you account for the vastness of GM’s portfolio?

So it goes back to our opening conversation about functional orientation. So my primary partner in crime is Dave Richardson, and he oversees our engineering team as well as our IT. That’s a very important part because, like I said, they’re the imagineers. We need them and the designers to turn this into reality. One big change that happened is we’re now a primary peer group to our counterparts on the vehicle product development side. That’s actually immensely freeing because now you have the people who are imagining the vehicle because here, I use this word internally too, we’re in “service” of the vehicle. That’s what we do. Software alone is not the goal. The goal is to create an amazing vehicle. You get into it and you take that ZR1 on the track, and our president Mark Reuss took this thing up to over 230 miles per hour.

There’s so much going on in that car to make that happen, but it means all of us came together to make that happen. That’s one big structural change, that software and services are now in a primary seat at the table alongside people who are building and designing the vehicles. The other thing is we’re actively looking into how we can be leaner, more efficient, and optimized for speed because there’s this debate that takes place: How big is too big when you try to move fast? Eventually, the node connections in that network — the collaboration tax — becomes greater than the benefit of adding one more person to the nodes. That’s the thing that we always actively ask ourselves: How do we make sure this is a lean winning team that is the right size with the right players and has the right amount of connections? Not too much, not too little.

But we also want to be in a place where we say fewer yeses and make great ideas happen rather than say a lot of yeses and struggle in timing, scheduling, and other things. So a new perspective we’re bringing into the mix is one of the things that may be different. What it takes to build great software and services might be a little bit different, requiring different modus operandi than maybe some of the other parts.

You came into a structure, you changed the structure, and you have reduced the size of the structure in the interest of speed. What is the biggest difference between the structure you came into and the structure you have now?

One is just functional orientation. The second area is — and this is one of my strong beliefs — that I believe leading experts is a great model. You really want to be the experts of at least most of the things you’re overseeing. Therefore, instead of generalists managing groups, we are actively hiring or finding experts within GM to lead the expert subgroup of their domain because then there’s this really healthy debate environment, call it dialectic, seeking truth. When you’re experts at something, you have this desire to always improve and question, but when you have generalists overseeing things, they can maybe make you generally efficient, but they’re not going to be equipped to question the details.

So you have leading experts and then you have the other things we talked about, which is how do you create the trust environment where these experts trust their peer experts? They go, “You know what? I’m going to defer to you Nilay. You are better at communications and how to do better storytelling, how to make things pithier. Tell me how you think we should tell that story?” I could be dangerous and say, “Hey, I’ve been doing this for three decades, I know how to do PR.” But why should I do that? I have some of the best communications people working with me to advise me and to help me achieve that goal. I don’t think experts leading experts and experts trusting each other and passing the ball are new to GM per se, but we are iterating on those things and how we want to operate as a software and services group.

GM is a giant company. I think you report directly to CEO Mary Barra, the CEO.

Yes.

But she has to oversee how the vehicle platforms are expressed, from GMC trucks to Chevy trucks to Buick sedans, to the whole litany of GM’s cars, platforms, and brands. How do you break the tie if Cadillac wants a software feature, but Buick doesn’t?

I’ll start with, it’s hard,” but I call it “It’s hard work. Let’s do it. There’s definitely a recognition that we have a capacity. Just like any product group and engineering group, you need to be honest about the capacity you have. You need to be thoughtful about what you can do and when in terms of quality. I always say to people, “You don’t just say no. You usually say when and how.” So, when you have such a vast portfolio, you need to be commensurately sized on the software side. We won’t build an amazing vehicle that we committed to just because the software group is too busy.

However, when you’re looking at your capacity, software is not the only capacity constraint. You need to be thoughtful about suppliers. Let’s say you want to do something really magical, and you need just the right camera or just the right sensor. Well, you need to also negotiate how you get that supply delivered to the plant, which also has a capacity constraint. I have great friends who help me navigate what it takes to build a vehicle, and we’re one of the constraints but also the enablers.

So our job is to figure out if we are sized right. Do we have the right people looking at the right problems? How do we manage our capacity over multiple years — not just one year?Vehicle portfolios are a bit more like venture capital investment-style funds that need to manage or optimize your portfolio rather than a hedge fund that’s hyperactively trading. When you start imagining a vehicle like the ZR1, those ideas started probably five, seven years ago because that’s what it takes to build a great car. Sometimes what you’re doing is just a new model year of an existing car. That’s a different task. Sometimes you’re doing a midlife cycle refresh of that car.It’s definitely a complex portfolio management task, but at the end, the products you end up building are so viscerally and emotionally fun. I just call it, “It’s hard work. Let’s do it.”

How do you break the ties though?

It’s a debate club, and what happens in debate clubs stays in debate club. So, the thing is we have such amazing leaders like Mary, Mark Reuss, leaders who work for Mark, some of my brethren like Ken Morris, Josh Tavel on the design side, Michael Simcoe. I mean, these people are legends in their own right in what it takes to build a multi-car, multi-brand high quality portfolio successfully and profitably. It’s a respectful debate club. I think there’s two options:. you either agree and commit or you disagree and commit. And you get to work.

This leads me right into the big Decoder question. You have a lot of decisions to make. What is your framework to make decisions?

I think I’ve already been sprinkling in some of the clues….

I can tell you’re a listener. I can always tell when the guest is a listener. You’ve been building your way here.

Look, this is a “started at home” kind of a moment, but it is true. I grew up in a household where questions were answered by questions. It was a very Socratic method of debate and dialectic. You sought truth, and you let the debate lead you to places with your own persuasions because we all have this visceral first reaction that we think we know. It’s very important for a leader to always catch yourself and go, “I actually don’t know many things about this. Let’s do data intuition.” There’s this really famous article I love. The headline goes “Data and Intuition: Good Decisions Need Both.” That’s what it is. It is both, but you debate it because decisions are rarely so neatly binary. They’re always on a spectrum of gray, and they’re rarely deeply wrong. They’re deeply right. You debate it, you weigh it, and then, finally, it comes to a single term: conviction.

At the end, you need to make a bet and you need to have conviction. Now, this is your moment of making a bit of reality happen. It’s easy to react to empirically observable reality because then everybody can observe the same thing and come to the same conclusion. Life is easy, right? You do research and you do. The hard part is that the decisions we’re making today are crystal balling what’s going to happen in five years in terms of taste, desirability, zeitgeist, and technology. The only way you get there is to hopefully, subjectively, objectively, make good judgment calls. You make a call, you have conviction, and you put your efforts behind making that reality because there’s a lot of things that are under your control. Let’s worry about those.

There’s this term that I picked up from my days at Amazon. It’s called controllable inputs. You worry about your controllable inputs. Other things are going to happen, and you have to react. With your controllable inputs, you make your decision and you commit and start executing. I always tell people that most of success comes from relentless iterative execution rather than some magical, amazing, brilliant light bulb idea. It’s incessant dedication and doing it over and over and over again. That commitment and conviction is what gets you there with decisions.

Okay, you opened the door with conviction, so I’m going to ask you this. This is what everybody has wanted me to ask you about. The big decision, which was made before you arrived at GM, was to drop smartphone projection like CarPlay and Android Auto from GM vehicles. I’m assuming that when you took the job at GM, you agreed with this decision. You certainly haven’t undone it in your year there. Why make that decision? Why drop CarPlay and Android Auto from GM vehicles?

Because there was a belief and a hypothesis, which I genuinely believe in, that we are best positioned and owe it to our customers to create the most deeply integrated experience that you can create with the vehicle. We are not shipping devices with just monitors; we’re not a monitor company. We’re building beautifully designed, complete thoughts and complete convictions. We say, “This car is designed to do the following things awesomely.” This is Silverado, this is what it stands for and this is what it does. Let’s get to it.

When you want to create something so seamless, it’s hard to think about getting into a car and going, “Okay, so I’m doing highway trailering, but let me flip to a totally different user interface to pick my podcast. By the way, it’s a single app-obsessed interface — it’s still hard to believe. So I pick my podcast, flip back to trailering. Oh, now I can also do Super Cruise trailering. Let me manage that. Then, wait, we’re now getting into potentially Level 3, Level 4 autonomy levels that should be deeply integrated with talking to the map where the lanes lie. But wait a minute, the map that I’m using doesn’t really talk to my car.”

As a product person, you’ll never do that to yourself because it’s literally like, “Oh my God, I made my life so hard to create amazingly seamless experiences.” At some point, you need to make that bold decision and say, “I am not going to try to accommodate and figure out how to make all of these work. I’m going to just burn the bridges and burn the ships and commit.” We are going to create a deeply vertical, harmonious experience that works across the vehicle that is optimized for my vehicle.

I can appreciate when you’re in that one-off rental car that you don’t want to be driving but is the only car that’s available when you land, be my guest. Use projection. But the car that I researched for months and decided to buy for its interior, exteriors, its propulsion, and every feature that it does for me? We bought a Colorado ZR2 Bison. We love cars in our family; as you can see, we’re very outdoorsy. We go gravel biking and rock climbing in Yosemite with my son. I want that car to be true and true optimized for what I bought it for, not some lowest common denominator moment. That’s why it was not an easy decision to embrace when I joined, I’ll be very honest with you. At first, it’s a seemingly hard decision and maybe unpopular decision, but it is the right decision because we are here building an end-to-end integrated product for you.

I just have to say this: a lot of times when I land at the airport and I get in the rental car, it’s a white Chevy Malibu. Are you going to leave smartphone projection alone in the fleet cars that get put in rental lots?

I’ll take the suggestion back to our senior leadership team. Let’s see what happens.

I’m just curious because you’re describing one very specific thing that people like smartphone projection for.

Yeah, you’re right.

It is a big market for GM. That’s the kind of fine-grained debate about smartphone projection I get from Verge readers and Decoder listeners. There are all these places where that idea sounds great, and there’s this whole list of other places where that sounds horrible actually. When I borrow my parents’ car, I do not want to live in their Google account. I just want to send my Spotify to the screen.

It seems like you’re burning the bridges, but you’re burning the bridges to a lot of other experiences that people have in cars.

The one thing that I believe and encourage our partners — and we work very closely with Apple, Google, Amazon, Spotify, and others — to believe is that some of those pieces of information and preferences should be accessible to other devices that you choose to use. I don’t think they need to be locked in this walled garden of a single domain. I personally believe I should be able to take my podcast preferences from one podcast app to another. I always find it interesting that some of these very generic preferences I declare are locked and loaded, especially for non-specific content.

I appreciate that a certain show is only on Netflix, let’s say — respect — but most song music catalogs, podcasts, mapping information, your home address, your work address, these are pretty generic pieces of information that you probably want to just enable in any vehicle you walk into to make that experience more personal. So that’s one thing that our team is also working on. How can we actually get that information to power any bespoke experience that’s in a vehicle rather than having to relinquish the full vertical experience to a totally different one?

So I just spent a weekend test-driving a Blazer EV. A lot of fun. That whole stack is built on Android, right? It’s Google services, the Play Store, apps. It wanted me to log into my Google account. That’s where all of that information came from. Then, it wanted me to log into my Spotify account, which is where a bunch of that information came from. At some point, you’ve kind of just put an Android tablet in the middle of my car. Why build on Android? What you’re describing is a big vision. and then the way it’s expressed right now is, “Well, here’s Android.”

The way I look at it is if we were to do a startup today, we would probably start playing with existing building blocks, right? If you and I were to do, let’s say, an overlanding app, you probably wouldn’t come to me and say, “I have an idea. First we should get a data center and get a bunch of racks.” We would probably go to AWS, and then you’d say, “Hey, for developing this app, we’re going to use React. There’s some existing open-source libraries, let’s take those.” And let’s say we were also in the business of using some kind of map. We would be like, “Well, when we’re working on this platform, we’ll use this map versus here where we don’t need to actually integrate a third-party map.” We’re not going to build our own map. Think about it that way.

As a software group and software developing group, we don’t have to build every single building block. You have capabilities out there that are already given to you. Android is a building block; it’s an open source system. In addition to that, Google provides building blocks that you can tap into and build around. We’re working with other companies and seeing if they’ll be interested in building their building blocks and their apps into this environment where we get to create the experience.

So, I look at this as just enabling capabilities and wanting these moments you talked about,how you need to log in here to get the information from that ecosystem and make it more personal. I’m going to challenge my team. I think we should make that super magical and simple. Why can’t we just detect that you have your phone in your pocket and do a very simple handshake instead of scanning a QR code? Today that’s the most seamless way of doing it, but there’s many different ways to do it.

The same thing with if you log in once, what if you log into one plus another GM built car? Can we carry that over? Can we create one authentication system around GM ID that carries some of your other credentials? All of these are, in my opinion, solvable problems, and we’re now setting out to do that. It takes dedication and having this platform thinking to be able to, one at a time and with dedication, remove all of these moments of friction you articulated. I agree with you, and we’re going to make it simpler and simpler.

One thing I’ll say though is, thankfully, most people don’t change their cars as often as you do. You’re changing your cars as often you probably change your shirts. Most people buy a car and love it for three, four years. I used to say this in other places I work, most people don’t upgrade the operating system of their phones every week with a new build. Most of us hold some of our devices for longer. Maybe you wish you didn’t have to do the setup routine, but you do it once and now it’s your car, it’s your vehicle, it’s your data, it’s your choices.

I want to come back to this timeline of three or four years. But first, I want to ask you a very challenging question — I will warn you, not every car executive who’s been on Decoder has been able to answer.

I’m listening.

So this is the challenge: Can you name all of Google’s car platforms and describe what they do?

Let me give you the phone number of a Google executive.

I don’t think they can do it either. I want to be very clear about this.

Look, I’ll say this, we have a great, tight partnership with them, and what I try to do when you look at our–

No, no, no, you’ve got to try.

I’m going to try.

Which ones do you use?

We definitely use Google built-in and Google Automotive Services. And this is the thing: most people don’t know macOS is actually built on a version of Unix. So when you actually start splitting the layers of the cakes, there is definitely open source or Linux, and Android is actually a Linux derivative, when you think about it. Let’s get all nerdy here. When you start looking at it, all of this is actually based on some other thing.

So we definitely use Google Built-In, and we use Google Automotive Services to create the experiences, but we create our own experience around it. With the Blazer EV, there are these moments to make that more personal. You can use Google Maps with all the things that you’ve done in Google Maps before. If you would like, you can sign in with your Google account so your homework or the last search you did on your phone also just appears in your recents. So those are the kind of moments where we hook into Google services. But yeah, it’s hard to build popular building blocks like Google does or AWS does and keep all their branding and naming right. So I hear you.

What’s the difference between Google Built-In and Android Automotive? Which is the Android that runs–

Android Automotive Services.

Well, there’s Google Automotive Services, which is basically the Play Store and the apps, and then there’s Google Built-In, which is what you have.

Correct.

And then there’s Android Automotive? I’ve been trying to keep this correct and clear in my head. I feel like you would know.

No, I do know.

You’re chipping the software–

I do know, I do know. What I’m laughing about is that consumers don’t need to know these things–

But I need to know. I’m dying to know. This is why you’re here.

So I’ll give you an answer that is going to be less satisfying than you want. We have different model years and different brands, some of them have been with us longer and some of them are brand new vehicles. You get into a Lyriq, you get a different experience. So our goal, first of all, is to unify those experiences. That’s why I’m here, that’s why Dave is here, that’s why all of our teams are here: to actually take away that complexity that you’re describing. You can actually see where in which vehicles we’re using different combinations of services, but again, I think it’s pretty opaque from a consumer perspective. I don’t think they need to know these differences, but I’ll point you to a great white paper that tries to explain it all.

Oh, I’ve read the white papers, I’ve seen the blog. I guarantee you more people are going to send me more blog posts. I’ll just keep asking. One day we’re going to figure this out.

What does the deal with Google to use its stuff look like? On a smartphone, that is the subject of pretty ferocious regulatory interests, litigation, you have to take Play services, you’ve got to use its app, and they get a 30% split. Is it the same on the car?

The commercial details of it are definitely confidential. I don’t think I should be in the business of disclosing our partnership details. But from our perspective — just like you articulated — it’s complex in that there are different types of experiences and building blocks, and depending on how you use those different blocks, there are different terms and conditions of how you integrate, operate, and update them.

So there is not a simple neat answer because this is unlike Windows. This is not a standard Windows PC where you buy a Windows license and you’re off to the races as an OEM. The cars we create are far more complex and capable, so therefore, a good portion of the cars are using things that are not directly licensed from Google while some portions are directly licensed for Google — where Google Maps is Google Maps or Google Assistant is Google Assistant. Every one of those permutations as they come together is different and unique.

The reason I ask that question is because one of the reasons so many car makers are interested in taking control away from smartphone projection and building their own user interfaces is what you have described: to offer more services, offer more subscription services in the car, offer more transactions on the screen, taking 30 percent of every transaction that happens on screen in the car, offering streaming services.

I’ve heard every version of these ideas, and if your economics include having to pay Google a fee for Android apps operating in your car on top of the fees you are extracting, it feels like there’s a real tension there, with Google as an operating system vendor or a services vendor wanting a fee or a percentage split on top of the one the GM wants from an app developer. And then the economics for an app developer to put an app on your car platform starts to fall apart pretty fast.

We see this play out in smartphones, right? I think the day before we were speaking, Hulu said, “we’re not going to allow sign-ups on iOS directly anymore because we don’t want to pay Apple a 30 percent split.” I see this tension on smartphones with just one intermediary in Apple and Google. In your car there’s two, right? There’s Apple, here’s Google, and then there’s potentially GM.

So let me parse this question a little bit. I’ll start with that our primary motivation in every conversation we have is to first create a great customer experience. I have this simple axiom: “great products also usually make great businesses.” But you always start with a great product. In everything we discuss, we’re always trying to continuously optimize how we create a great experience for you. That’s it, that’s the primary goal.

The second thing is the type of services we are building that I am really excited about. The good thing is that we are a company that has already proven that model with OnStar. We have millions of subscribers who are using it to extend their vehicle experience in a way that is expected from the vehicle, which is how do you make this safer for me? How do you help me recover my car if there was ever a theft? How do we help you insure your car easier because that is, again, integral to your car. Super Cruise is a great example, right? Super Cruise is ever improving. We continuously add more and more roads, for example, which means we’re adding a concrete value. We just upped that to 750,000 miles across the U.S. That means we’re investing, therefore there’s value added and value created, so consumers are willing to pay for that subscription.

Those are the types of services I’m deeply interested in: things that are extending my experience as a vehicle. I think that’s ours, that’s ours to build and ours to develop. Then there are some opportunities like you articulated around entertainment, around gaming, etc. But to me they come next. I really, and we are obsessed with extending the vehicle experiences first because we know how to do that really well and there are some domains.

I’m not here to compete with Spotify or Apple Music. They do a fantastic job servicing all of our music listening needs. We work with them, we bring their apps to our platform, and we run it. So I am not too concerned about that part that you articulated: “How do you create value? How do you capture value? How do you split the economics?” I think those are all solvable problems, but it all starts with the experiences I owe you as a consumer who buys a beautiful Escalade IQ in a month or two. What are the experiences I should create? When you get into that car, you’re going to see this pillar-to-pillar display with front control and back control. We have so many ideas of what experiences we can create for you there before we get into this area of revenue and shared debates around a gaming app.

Earlier, you mentioned three to four years, and the reason that immediately lit up my brain is because when listeners ask me about CarPlay, a thing they say to me is, “Well, if you’re going to put a giant Android tablet in my car, that means I have a giant Android tablet in my car and Android tablets get slow. Three years from now, that will be slow, and with smartphone projection, it’s always running off my phone, which is always fast.”

How do you deal with, “Okay, we’ve put all of the user interface on an SoC in a car that might have to last for a decade?” Do you just over-spec them to make sure they have enough headroom to last that long? Do you just assume people are going to turn them in on three-year leases? What’s the story?

You deal with multiple things. First of all, I’m a huge believer in creating elegant software. We’re not just simply running some off-the-shelf Android and just conceding the entire thing in a non-optimized way. We’re directly involved. We actually are using that as a building block and we’re creating our experience. Therefore, the onus is on us. How do we create more optimal and optimized code? I’m a huge believer that you can continuously improve your code from the day you ship till the day you update it. We are on a continuous optimization journey. This thing doesn’t stay stable and just collects on top of that additional debt.

In the other areas, you’re absolutely right. You do look at the compute investment you’re making and you want to give yourself headroom because today we’re doing Level 2 [autonomy], which is, as I call it, “eyes-on, hands-off driving” with Super Cruise. But we are also the company that has already invested in a fully autonomous future, which is with Level 4. There is a zone of Level 3 autonomy. Well, if you want to live on that journey, you want to create a system that has some headroom or is extensible and expandable as you add new capabilities. Maybe you don’t want to over-provision from day one because you’re not going to take advantage of it.

I never worked in this industry, but I am familiar with it because I was a gamer — I have less time to do gaming now — but the same thing applied when you were creating the next PS5 versus the PS6. You knew this platform was going to be there for many years to come, and you were trying to create just the right balance of provision with room for extendability, optimized software, and bringing content in that feels fresh, new, and always improving. That’s exactly what we’re doing with our vehicles. If you bought a Lyriq last year, the Lyriqyou’re driving today has newer software, is more optimized, and probably running faster than when you bought it because we’re continuously improving it.

You mentioned other platforms. I always think about Windows Phone for some reason when people talk about platforms like this. Again, when listeners tell me why they like CarPlay, which they’re constantly telling me about why they like CarPlay, one of the things they say is, “All of the apps on my phone support CarPlay.” I have one listener who told me he routinely switches between six and seven apps in CarPlay while he’s commuting. It’s like podcast apps, ebook apps, all this stuff, and he’s just flipping. I find that incredible. I use two apps, but so it goes.

That means you need the app support. If you want to sell that listener a Blazer or a Lyriq, he’s got to sit down, open the Play Store — which is not great for discovery right now, I’ll tell you — and he’s got to search for all of his seven apps and he’s got to find them all, which is the problem Windows Phone had, right? The volume of apps wasn’t there, and the library wasn’t there. Have you thought about solving that problem? Is that GM’s problem? Is that Google’s problem? How are you addressing that?

I think first of all, anything that happens in our cars is our problem. That’s how I look at it. There’s that finding that happy medium between what are the most popular apps we believe should be there to complete your experience, and sometimes those ones — and this even applies today — that don’t even have apps. Even with your phone you sometimes need to go punch through to the internet and go to the browser and do it, right? These types of different domains exist in our lives in other places too. We are working actively with the people we believe provide the core apps, maybe the top 10. It’s even more stark when you start talking to people. What are the must-have apps while you are driving or while you are a passenger when you’re riding in our vehicles? That list is actually far shorter than the long tail of things we do.

I think you’re touching up on an interesting area that I am also interested in, which is if there is a way to bring vehicle-optimized versions of certain apps. Let’s say you’re a surfer, there’s this great app called Surfline. Wouldn’t it be fascinating if you had a vehicle-optimized version of the Surfline app that gave you information about what the tide conditions are, wind, etc. but in a smart way? Now, let’s go a little bit into the future. I believe this rabbit hole of vertical, micro-segmented apps is going to go under a change too, especially with voice interfaces becoming far more intuitive than ever before because of large language models.

For me, I think the biggest change with large language models is that I can finally be understood by the machine despite the fact I did not phrase myself perfectly. That device is going to be able to interpret my intent and give me answers. The reason I think we all default to a more touch user interface today is that we as users have to do a lot of the disambiguation instead of the machine understanding what you want. You are going through things, when all you want to know actually is this. You don’t need an app. If you want to drive to Aptos, California, to surf in an hour: “Tell me the surf conditions.” It’s a two-sentence answer. Do I really need the app for that?

Apple’s solution to this is next-generation CarPlay where it will send more information to more screens, where it will help car makers design their center console, help car makers design the instrument cluster and other screens. Your old boss, [Greg Joswiak], is a big advocate of this. It’s been taken up a little bit, but not a lot. Have you looked at it? Would you use next-gen CarPlay?

I look at almost everything. I’ve driven almost every competitor’s car under the sun, and I have all the devices that you probably also carry in your pocket to learn, test, and understand. We already made our decision about the experience we want to create, which is going to be a deeply vertically-integrated experience that GM designs, builds, maintains, updates, and innovates on for our customers. That’s our direction.

There’s a really interesting A/B test happening in the market right now. You’ve got the Blazer EV, which has your approach. There’s the Honda Prologue, which is substantially the same platform, and the Acura ZDX, which is substantially the same EV platform. They support CarPlay. The Prologue is kind of a hit, I think a surprise hit for Honda. Are you keeping track of that? Do you think that’s attributable to CarPlay? Is that difference in branding?

Like I said, we’re looking at the competition, but we’re probably a customer and quality-obsessed company rather than competition-obsessed company. We’re definitely a highly competitive company, but it goes back to something we talked about: It’s about conviction. We have high conviction that this is the right path for us to take to create truly interior experiences for our customers. Like I said, an easier decision would be to ask, “Why make that effort?”

But we have a strong conviction that effort pays off in a better customer experience. You get the most out of your vehicle because now we’re the company that builds the vehicle and is also creating the infotainment experience, the cluster experience, the app, and everything. We’re going to build that one day and maybe a voice assistant on top of it. The only way you can create that end-to-end magic is to have a strong conviction that you want to own all of these. It’s hard to create those seamless experiences. You’re going to always feel the seams, and I don’t think you should be in the business of feeling the seams when you buy a $30,000 to $100,000 car. It’s our job to make it beautiful and seamless.

We have a lot of listeners who say, “Look, I’m just not going to buy a car without CarPlay, and that’s fine.” Point them to something that will make them change their mind, that will make them reconsider. What’s something coming up where you’re going to say, “I can build you an experience that will make you reconsider?”

I don’t think you need to wait for the future. I think Super Cruise and how it integrates natively with the–

You can get Super Cruise on a [Cadillac] Lyriq with CarPlay today.

Well, the thing is that when you are using the native integration with the map that knows where you’re going to be, what are the road conditions, what’s happening, where are the construction sites, and it ties it back to the battery system and says, “By the time you get to your destination and return, your battery’s going to be at this level. Here is the energy efficient route you can take.”

I cannot create those magical little dots connecting to each other if I had to go back and forth between a navigation app running on another platform, and that’s the thing: you don’t need to wait for the future to create that. The other things that we want to create are similar things, especially when we start introducing Level 3 autonomy, hands off, eyes off, and even with robotaxis. That type of integration is a prerequisite, not an optional feature.

Well, Baris, thank you so much for coming on Decoder. You got to come back so we can do a full hour on robotaxis, and a full hour of the difference between product and program management.

Nilay, I really enjoyed our conversation. Great questions, and I’m going to listen to this episode too. I hope it turns out pretty great.

We can all learn something. Maybe we’ll get that blog post from Google. I’ll let you know. Thank you, Baris.

Thanks so much. Take care.

Valorant is winning the war against PC gaming cheaters

Image: Alex Parkin / The Verge

Riot Games’ investment into its Vanguard system is paying off. Multiplayer games on PC were a mess back in 2020. Developers were struggling to respond to blatant cheating as more and more people turned to gaming at home during the covid-19 lockdowns. Call of Duty: Warzone, PUBG, and Destiny 2 were all riddled with people using aimbots to automatically shoot opponents or wallhacks to see everyone on a map.
Riot Games’ Valorant stood out because of its controversial and aggressive anti-cheat system, Vanguard, which had the potential to keep cheaters away. Now, four years later, it’s clear that Vanguard is winning the war against PC cheaters unlike any other anti-cheat system.
“We don’t see as many of the cheats that try to function on the machine and get access,” says Phillip Koskinas, director of anti-cheat on Valorant, in an interview with The Verge. “That has just become too much of a chore for cheat developers.”
Vanguard has made it far more difficult for PC gamers to use things like aimbots or wallhacks. This is partly due to a controversial kernel-level driver that is always running after you boot your PC. Riot’s Nick “Everdox” Peterson developed a system in Vanguard that detects when cheat engines are trying to get access to Valorant. “He came up with a fairly novel way to know that something has been mapped into kernel memory that isn’t supposed to be there,” says Koskinas. “The method is so cute that I can’t explain it because they’ll figure it out too quickly.”
The method sounds like it works similarly to when you crack open a piece of hardware and those little plastic clips fall off to let the device manufacturer know you have voided the warranty. “Once that’s done, we know that something happened and then we just wait to see something occur on Valorant that confirms you’re using it for cheating,” says Koskinas.
That’s led cheaters to move increasingly toward hardware to bypass systems. One of the most popular ways that cheat engines now hook into games involves direct memory access (DMA) with dedicated hardware. “You’re basically using a PCIe card to request reads of physical memory,” explains Koskinas. “They have developed techniques with these cards, the most popular one being Squirrel, to do a lot of traditional memory scanning but totally externally.”
That means a cheater will have a secondary PC that is scanning the memory space of Valorant, looking for player positions. A cheater can use this second PC with a monitor to display a special new radar that lets them know exactly where opponents are. It’s a devastating cheat in a game like Valorant, where players rely on tactics, positioning, and stealthiness to get an advantage.

Image: Riot Games
DMA cheating involves dedicated hardware.

Riot has also developed methods to detect this new form of hardware-level DMA cheating thanks to Peterson. His invention essentially blocks reads to internal memory by suspicious devices. I recently ran into an issue with this DMA protection, as Vanguard started blocking my network card every time I loaded into a Valorant game. Riot has a list of hardware and firmware that is trusted, but the network card on my motherboard was using a method that looked suspicious. The issue was rectified within hours, but it showed how powerful Vanguard was that it could knock out my PC connectivity until I rebooted.
Most of the cheats for Valorant these days have been reduced to triggerbots, programs that use screen readers to look at the center of your monitor and then automatically shoot when a player’s crosshair is placed over an enemy. Koskinas says these account for “about 80 percent” of cheats in the game.
The addition of Vanguard to League of Legends earlier this year also dramatically reduced scripters, and the League team revealed in August that it had banned more than 175,000 accounts for cheating since Vanguard was introduced.
That’s encouraging for Valorant and League, but the situation isn’t as bright for other game developers that build their own anti-cheat systems. A recent study from the University of Birmingham revealed that cheats for Activision’s Call of Duty: Warzone remain accessible and affordable, and that Activision’s Ricochet anti-cheat falls short against more sophisticated cheats. Activision even had to fix an anti-cheat hack in Warzone and Modern Warfare III that led to legitimate players getting banned.
“Ricochet has talented individuals on the team, but they clearly do not have enough funding or freedom,” says zebleer, the developer behind Phantom Overlay — one of the most popular cheat engines for games like Call of Duty, Overwatch 2, and more. “Call of Duty is overrun with cheaters. They are implementing quick fixes. They are not implementing things they should be implementing likely because Activision won’t let them.”
Zebleer thinks Vanguard is clearly winning against cheaters, thanks to the anti-cheat team having funding, talent, and freedom. Riot has hired engineers that have developed cheat engines in the past, including Koskinas, who developed and sold cheats more than 15 years ago to help fund his academic career.
Unsurprisingly, the researchers at the University of Birmingham agree that Valorant has the best anti-cheat system. It was ranked at the top of the anti-cheat pile, followed by Fortnite, which also uses a kernel-level system. Counter-Strike 2, Battlefield 1, and Team Fortress 2 were ranked at the bottom.
The researchers also highlighted weaknesses in Windows protections that allow cheat software to inject itself into the kernel, just like malware does. After the devastating CrowdStrike incident, Windows kernel access has become a hot topic as Microsoft is increasingly looking at ways to help CrowdStrike and other security vendors operate outside of the Windows kernel.
Riot is looking to Microsoft to help secure Valorant further. “Microsoft got a lot more proactive about revoking the certificates for drivers that were malicious,” says Koskinas. “We kind of chase what Windows is willing to do, so if they start requiring virtualization-based security to be on, or hardware-enforced stack protection, or hypervisor code integrity, we will leverage those features that protect Windows for us and just require them to be on and recede from the kernel space.”
Vanguard will soon only start when the game launches, provided you’re using all of the latest Windows 11 security features, instead of being always-on after boot. That should help with some of the privacy concerns, too.
Riot’s focus for anti-cheat is on Windows right now, and there are no plans for Linux support with Valorant or League of Legends. While the Steam Deck supports some anti-cheats, developers like Riot are increasingly shying away from Linux. “You can freely manipulate the kernel, and there’s no user mode calls to attest that it’s even genuine,” says Koskinas. “You could make a Linux distribution that’s purpose-built for cheating and we’d be smoked.”
Respawn just dropped support in Apex Legends, citing similar concerns to Riot about cheating. Epic Games also refuses to support Fortnite on Steam Deck / Linux due to a lack of users. “Imagine if Steam Deck just has the security handled so we know it’s a genuine device, it’s fully attested, all these features are enabled, we’d be like cool, go game, no problem,” says Koskinas.
While Riot seems to be on top of traditional PC cheating, it may have to contend with AI-powered cheating soon. That could come from dedicated hardware like MSI’s monitor that helps you cheat in League of Legends or screen readers that get increasingly complex. Riot is particularly concerned with image reading. “That is where all cheating is heading,” says Koskinas. “We’ve done a lot of research into what human mouse and keyboard input looks like, but it is a concern.”
One possible future could see AI cheats and AI detection battling against each other in a virtual war. “We’re at a disadvantage, honestly. [AI models] can learn what human input looks like,” says Koskinas. Valorant is winning the war right now, but AI could reset the playing field of this ongoing cat-and-mouse game.

Image: Alex Parkin / The Verge

Riot Games’ investment into its Vanguard system is paying off.

Multiplayer games on PC were a mess back in 2020. Developers were struggling to respond to blatant cheating as more and more people turned to gaming at home during the covid-19 lockdowns. Call of Duty: Warzone, PUBG, and Destiny 2 were all riddled with people using aimbots to automatically shoot opponents or wallhacks to see everyone on a map.

Riot Games’ Valorant stood out because of its controversial and aggressive anti-cheat system, Vanguard, which had the potential to keep cheaters away. Now, four years later, it’s clear that Vanguard is winning the war against PC cheaters unlike any other anti-cheat system.

“We don’t see as many of the cheats that try to function on the machine and get access,” says Phillip Koskinas, director of anti-cheat on Valorant, in an interview with The Verge. “That has just become too much of a chore for cheat developers.”

Vanguard has made it far more difficult for PC gamers to use things like aimbots or wallhacks. This is partly due to a controversial kernel-level driver that is always running after you boot your PC. Riot’s Nick “Everdox” Peterson developed a system in Vanguard that detects when cheat engines are trying to get access to Valorant. “He came up with a fairly novel way to know that something has been mapped into kernel memory that isn’t supposed to be there,” says Koskinas. “The method is so cute that I can’t explain it because they’ll figure it out too quickly.”

The method sounds like it works similarly to when you crack open a piece of hardware and those little plastic clips fall off to let the device manufacturer know you have voided the warranty. “Once that’s done, we know that something happened and then we just wait to see something occur on Valorant that confirms you’re using it for cheating,” says Koskinas.

That’s led cheaters to move increasingly toward hardware to bypass systems. One of the most popular ways that cheat engines now hook into games involves direct memory access (DMA) with dedicated hardware. “You’re basically using a PCIe card to request reads of physical memory,” explains Koskinas. “They have developed techniques with these cards, the most popular one being Squirrel, to do a lot of traditional memory scanning but totally externally.”

That means a cheater will have a secondary PC that is scanning the memory space of Valorant, looking for player positions. A cheater can use this second PC with a monitor to display a special new radar that lets them know exactly where opponents are. It’s a devastating cheat in a game like Valorant, where players rely on tactics, positioning, and stealthiness to get an advantage.

Image: Riot Games
DMA cheating involves dedicated hardware.

Most of the cheats for Valorant these days have been reduced to triggerbots, programs that use screen readers to look at the center of your monitor and then automatically shoot when a player’s crosshair is placed over an enemy. Koskinas says these account for “about 80 percent” of cheats in the game.

The addition of Vanguard to League of Legends earlier this year also dramatically reduced scripters, and the League team revealed in August that it had banned more than 175,000 accounts for cheating since Vanguard was introduced.

That’s encouraging for Valorant and League, but the situation isn’t as bright for other game developers that build their own anti-cheat systems. A recent study from the University of Birmingham revealed that cheats for Activision’s Call of Duty: Warzone remain accessible and affordable, and that Activision’s Ricochet anti-cheat falls short against more sophisticated cheats. Activision even had to fix an anti-cheat hack in Warzone and Modern Warfare III that led to legitimate players getting banned.

“Ricochet has talented individuals on the team, but they clearly do not have enough funding or freedom,” says zebleer, the developer behind Phantom Overlay — one of the most popular cheat engines for games like Call of Duty, Overwatch 2, and more. “Call of Duty is overrun with cheaters. They are implementing quick fixes. They are not implementing things they should be implementing likely because Activision won’t let them.”

Zebleer thinks Vanguard is clearly winning against cheaters, thanks to the anti-cheat team having funding, talent, and freedom. Riot has hired engineers that have developed cheat engines in the past, including Koskinas, who developed and sold cheats more than 15 years ago to help fund his academic career.

Unsurprisingly, the researchers at the University of Birmingham agree that Valorant has the best anti-cheat system. It was ranked at the top of the anti-cheat pile, followed by Fortnite, which also uses a kernel-level system. Counter-Strike 2, Battlefield 1, and Team Fortress 2 were ranked at the bottom.

The researchers also highlighted weaknesses in Windows protections that allow cheat software to inject itself into the kernel, just like malware does. After the devastating CrowdStrike incident, Windows kernel access has become a hot topic as Microsoft is increasingly looking at ways to help CrowdStrike and other security vendors operate outside of the Windows kernel.

Riot is looking to Microsoft to help secure Valorant further. “Microsoft got a lot more proactive about revoking the certificates for drivers that were malicious,” says Koskinas. “We kind of chase what Windows is willing to do, so if they start requiring virtualization-based security to be on, or hardware-enforced stack protection, or hypervisor code integrity, we will leverage those features that protect Windows for us and just require them to be on and recede from the kernel space.”

Vanguard will soon only start when the game launches, provided you’re using all of the latest Windows 11 security features, instead of being always-on after boot. That should help with some of the privacy concerns, too.

Riot’s focus for anti-cheat is on Windows right now, and there are no plans for Linux support with Valorant or League of Legends. While the Steam Deck supports some anti-cheats, developers like Riot are increasingly shying away from Linux. “You can freely manipulate the kernel, and there’s no user mode calls to attest that it’s even genuine,” says Koskinas. “You could make a Linux distribution that’s purpose-built for cheating and we’d be smoked.”

Respawn just dropped support in Apex Legends, citing similar concerns to Riot about cheating. Epic Games also refuses to support Fortnite on Steam Deck / Linux due to a lack of users. “Imagine if Steam Deck just has the security handled so we know it’s a genuine device, it’s fully attested, all these features are enabled, we’d be like cool, go game, no problem,” says Koskinas.

While Riot seems to be on top of traditional PC cheating, it may have to contend with AI-powered cheating soon. That could come from dedicated hardware like MSI’s monitor that helps you cheat in League of Legends or screen readers that get increasingly complex. Riot is particularly concerned with image reading. “That is where all cheating is heading,” says Koskinas. “We’ve done a lot of research into what human mouse and keyboard input looks like, but it is a concern.”

One possible future could see AI cheats and AI detection battling against each other in a virtual war. “We’re at a disadvantage, honestly. [AI models] can learn what human input looks like,” says Koskinas. Valorant is winning the war right now, but AI could reset the playing field of this ongoing cat-and-mouse game.

Oops, Apple approved another illegal streaming app

The French-language app shows new movies like Terrifier 3 and Venom: The Last Dance. | Image: Thomas Ricker / The Verge

Another illegal streaming app has made its way to the App Store — but it only surfaces pirated films for people in certain regions outside the US, including France, Canada, and the Netherlands.
As shown in a post on Threads, the App Store listing for “Univer Note” presents itself as a productivity platform that can “easily help you record every day’s events and plan your time.” However, if you’re a user in certain countries, like France or Canada, opening the app shows a collection of pirated movies, such as Venom: The Last Dance, Joker: Folie à Deux, and Terrifier 3. Options within the app are labeled in French, while films stream in their original language with French subtitles or dubbing.
Anyone else who downloads the app in an unsupported region, such as the US, will only see a productivity app, a strategy we’ve seen used by other piracy apps to evade detection from reviewers. One recent example was Kimi earlier this year, which posed as a vision-testing tool and was quickly removed after news outlets began reporting about it. Apple didn’t immediately respond to a request for comment about this app.

The French-language app shows new movies like Terrifier 3 and Venom: The Last Dance. | Image: Thomas Ricker / The Verge

Another illegal streaming app has made its way to the App Store — but it only surfaces pirated films for people in certain regions outside the US, including France, Canada, and the Netherlands.

As shown in a post on Threads, the App Store listing for “Univer Note” presents itself as a productivity platform that can “easily help you record every day’s events and plan your time.” However, if you’re a user in certain countries, like France or Canada, opening the app shows a collection of pirated movies, such as Venom: The Last Dance, Joker: Folie à Deux, and Terrifier 3. Options within the app are labeled in French, while films stream in their original language with French subtitles or dubbing.

Anyone else who downloads the app in an unsupported region, such as the US, will only see a productivity app, a strategy we’ve seen used by other piracy apps to evade detection from reviewers. One recent example was Kimi earlier this year, which posed as a vision-testing tool and was quickly removed after news outlets began reporting about it. Apple didn’t immediately respond to a request for comment about this app.

Recent Posts

Recent Comments

New Model Calculates Chances of Intelligent Beings In Our Universe and Beyond

Prioritize Your Mental Health and Nutrition This Thanksgiving With These Tips

Bpftune uses BPF to auto-tune Linux systems

Categories

Archives

Recent Posts

Recent Comments

New Model Calculates Chances of Intelligent Beings In Our Universe and Beyond

Prioritize Your Mental Health and Nutrition This Thanksgiving With These Tips

Bpftune uses BPF to auto-tune Linux systems

Categories

Archives

Prime Video will let you summon AI to recap what you’re watching

Elon Musk’s PAC admits $1 million voter giveaways aren’t ‘random’

Amazon’s Echo Spot alarm clock is on sale with a free color smart bulb

Read our Echo Spot review.

A few other ways to save

Can Philadelphia’s ballot counters outrun election lies?

PS5 Pro’s actual specs and the launch day list of upgraded games revealed

X is allowing people you’ve blocked to see your posts

The New York Times Tech Guild is on strike

GM software boss: we have ‘high conviction’ ditching CarPlay is the right path

Valorant is winning the war against PC gaming cheaters

Oops, Apple approved another illegal streaming app