The Most Interesting Thing in AI

However you feel about Meta, Andrew Bosworth has likely played a part in it. The long-serving right hand to Mark Zuckerberg and current Chief Technology Officer, “Boz” has helped shape almost every stage of the company’s growth. He’s credited with introducing the Facebook News Feed in 2006, and, more recently, with overseeing the company’s efforts in the metaverse.

Now, Boz is a key figure behind Meta's artificial intelligence strategy. In late June, he joined Nicholas Thompson, CEO of The Atlantic, for a conversation covering the state of Meta AI: from the privacy concerns around its glasses to its controversial (and now halted) program training models on its employees’ keystrokes and mouse movements to its pivot from open-source AI to a proprietary model.

(00:00) Introduction

(02:34) Can AI glasses get past the developer ecosystem Catch-22?

(05:45) What’s true and false about NameTag, facial recognition, and Meta’s glasses

(09:22) Why Meta tracked employee keystrokes to teach AI how to use computers—and why it stopped

(12:20) The legal complications of tracking employee data

(15:58) Why unique professional workflows are more valuable than generic internet text

(18:15) Boz’s vision for AI: a tool to maximize individual human potential, not replace workers

(21:15) How AI produces faster coding but slower thinking, leading to "well-executed bad products"

(24:10) If AI doubles a worker's productivity, do companies hire fewer people or double the team size?

(28:38) Why Boz is optimistic about life "under the algorithm"

(32:08) Meta’s pivot from open source (Llama) to closed models (Muse Spark)

(37:25) How long will scaling laws hold?

(40:26) Where Meta already excels in AI

(46:27) Why energy production is the upstream constraint for the future of AI
Learn more about your ad choices. Visit megaphone.fm/adchoices

AI agents can scour the internet for us, reply to our messages, and add events to our calendars. But in order to do so, they need sweeping access to our data. Handing that over presents an unprecedented threat to privacy, says Signal Foundation president Meredith Whittaker, who directs the Signal messaging app. In conversation with The Atlantic CEO Nicholas Thompson, Whittaker discusses the risks of agentic AI, Signal’s approach to AI coding, and why it’s uncompromising in its commitment to protecting user data. 

This episode was recorded prior to Keir Starmer’s resignation announcement.

(00:00) Introduction

(02:52) AI’s threat to privacy

(03:28) The end of the operating system as a safe boundary

(04:15) The dangers of off-device compute and the limits of Apple’s privacy promises

(05:30) Prompt injection vulnerabilities, and the risks of giving agents access to calendars, contacts, messages

(06:50) How Signal mitigates the risk of AI software like Microsoft Recall

(09:15) The accessibility trade-off: Why protecting privacy (at this moment) comes at the cost of screen readers for blind/low-vision users

(12:20) Are AI agents the new root users?

(15:28) The problem of proprietary OS providers

(17:40) Why don’t consumers care about privacy?

(19:35) Can a focus on enterprise customers fix AI’s privacy problems?

(23:30) How government surveillance can undermine private industry privacy efforts

(28:13) How Signal worked with Apple to fix the iOS notification exploit

(30:50) Meredith Whittaker’s critiques of the UK government’s content-scanning proposal

(36:20) The risks of compromising on privacy

(42:30) Would Signal take a different approach to privacy if it were the world's largest messenger app?

(45:36) What would you do with unlimited funding?
Learn more about your ad choices. Visit megaphone.fm/adchoices

Take a scroll through Facebook or TikTok today, and you’ll likely find some AI-generated content: an influencer with impossibly-radiant skin, a stunt that bends the laws of physics, Shrimp Jesus. And while there might not be much social benefit to AI slop, that doesn’t mean that we should shun AI from our feeds entirely, says Eli Pariser. As the author of “The Filter Bubble,” Pariser has been a longtime critic of social media. But he’s also the cofounder of Upworthy, the bubbly bright news site that perfected the clickbait headline. Today, Pariser is building Roundabout, a kind of socially-conscious answer to Nextdoor, and he’s hopeful that AI agents can mend the thumb-shaped hole in our social fabric. But is more tech really the answer to our tech woes? Nicholas Thompson, CEO of The Atlantic, sits down with Eli to find out.

(00:00) Introduction

(02:00) How we can avoid social media's worst mistakes

(03:36) Would subscription models have fixed social media?

(05:03) How the agentic interface will change news consumption

(07:19) If AI replaces social media, where is the new digital town square?

(09:29) Nextdoor, engagement, and online utility

(11:42) Agents replace utilitarian uses; humans stay in trusted spaces

(13:37) The future of group chats: custom micro-platforms with add-on features

(21:30) Can a social network built around community stewards and offline events thrive? The case for Roundabout

(27:26) Upworthy grew fast but VC pressure drove it to clickbait

(33:45) AI agents = filter bubbles on steroids with deep personalization

(39:45) Fund external AI governance structures rather than internal alignment

Learn more about your ad choices. Visit megaphone.fm/adchoices

You've probably heard it already: AI is going to radically change the way we work. But the details change depending on who's making the prediction: AI will wipe out the C-suite, or entry-level jobs, or make us all into prompt engineers. Those scenarios are far-fetched, says Dan Priest. In his role as the Chief AI Officer at PwC, Priest sees firsthand how companies across the industrial landscape are utilizing AI– often in ways that clash with those  fatalistic prognostications. In a vital conversation with Nicholas Thompson, CEO of The Atlantic, Priest unpacks what he's learned about AI implementation, agents, and how businesses can adapt in the age of AI.

(00:00) Introduction

(02:30) The two extremes of AI adoption

(04:46) Will we see a billion-dollar one-person company?

(06:24) Why technology advances faster than implementation 

(07:32) Where is AI actually being deployed? From IT/code generation to CFO/CHRO functions

(10:17) Specialist vs. generalist roles in an AI world 

(12:17) Should every employee have their own agent? 

(14:39) Agent performance limits: Task length, concentration windows, and why multi-model checks and balances matter

(18:11) Who gains most from AI: top performers or early career workers?

(21:37) Agentic applications 

(26:12) The hourglass organization: Replacing pyramid and diamond models with expanded entry points plus leadership, compressing middle management

Learn more about your ad choices. Visit megaphone.fm/adchoices

In the last year, AI has arguably made more progress in coding than in any other domain. Its technical capabilities – harnessed through apps like Codex and Claude Code – have changed the way engineers work. Boris Cherny, Anthropic’s head of Claude Code, claims that 100% of his output is now generated by AI. But what does the rapid advance in coding tools mean for engineers and businesses?

To answer that, Nicholas Thompson, CEO of The Atlantic, sits down with engineer, author, and entrepreneur Paul Ford. The two discuss how AI has changed his work, its implications for a range of industries, and their missed opportunity to start a billion-dollar business (maybe).

(00:00) Introduction

(01:53) Writing vs. Engineering: Which profession has been transformed more by AI?

(03:23) The pitfalls of AI writing and the limits of AI agents

(04:09) How AI has changed the code development business model

(09:01) "Vibe coding": Paul Ford's personal experience building products through AI prompts

(10:47) "Death is coming" - The feeling of watching entire industries dissolve

(12:54) Where consensus programming thrives under AI assistance, and where it struggles

(15:18) Why making good software remains difficult despite faster code generation

(17:05) The restaurant review problem: The experiential engineering tasks that AI can't replicate

(21:33) Can you ship AI code without a human reviewing it?

(23:01) Will AI become self-recursive? How code runs, finds bugs, and self-corrects through iteration

(26:58) Engineers can produce 50x more code than before. What does that mean for PMs and design teams?

(29:46) Velocity vs. quality: Can AI genuinely upskill mid-level talent or just increase output speed?

(35:32) Should you teach your kids to code?

(38:36) What we learned from not building Grammarly

(42:04) Beyond agents: Incremental improvements and classic software-in-the-loop hybrid approaches
Learn more about your ad choices. Visit megaphone.fm/adchoices