Welcome back! Ronan Farrow just published a 100-interview investigation alleging a career-long pattern of deception at the top of OpenAI, Anthropic's revenue run rate has tripled to $30B and overtaken its rival, three of the biggest CEOs in corporate America all cited AI as the reason they are stepping aside, and Sam Altman published a 13-page blueprint asking the US government to prepare for superintelligence.
In today’s Generative AI Newsletter:
The New Yorker: What do 100 interviews and unseen memos reveal about Sam Altman?
Anthropic: How did Claude's maker triple revenue and overtake OpenAI in four months?
Fortune 500: Why are the CEOs of Walmart, Coca-Cola and Adobe all stepping down at once?
OpenAI: What does a "social contract" for superintelligence actually look like?
Latest Developments
The New Yorker's Altman Investigation Is the Most Detailed Portrait of OpenAI's CEO Yet
Ronan Farrow's investigation draws on more than 100 interviews, previously unseen memos from former chief scientist Ilya Sutskever and years of private notes kept by Anthropic CEO Dario Amodei. It is the most thoroughly sourced piece of reporting on Altman's leadership to date.
Sutskever's memos: Compiled from 70 pages of Slack messages and HR documents, they allege Altman misrepresented safety protocols to the board. One memo reportedly begins with a list headed "Sam exhibits a consistent pattern of ..." with the first item being "Lying."
Amodei's notes: Kept privately for years, they reach the same conclusion independently. The fundamental problem at OpenAI, Amodei wrote, was Altman himself.
The Microsoft comparison: A Microsoft executive told the reporters there is a small but real chance Altman will be remembered alongside Madoff and Bankman-Fried.
Career-long scope: The reporting spans conflicts at his first startup Loopt, Y Combinator partners allegedly trying to push him out, and the board drama that briefly ousted him in 2023.
Eroded safeguards: The board empowered to fire the CEO has been filled with allies. The independent investigation into the allegations behind his attempted ousting never produced a written report. Altman told the New Yorker his "vibes don't match a lot of the traditional AI-safety stuff."
There is no single smoking gun. There is a deeply sourced pattern spanning two decades, published at the exact moment Altman is asking the world to trust him with superintelligence. For every detractor quoted, there are fiercely loyal supporters. But this is now firmly in the public record.
Special highlight from our network
Build faster with Claude Code, without guessing what works.
In 6 live sessions, you’ll build your own claude.md system and learn a clear 13-step workflow to go from idea to shipped code. Use real examples, reusable skills, and proven patterns across development, testing, and docs.
No theory-heavy lectures. No prompt guessing.
Just a system you can use from day one.
Special highlight from our network
Right now, a lot of people feel this pressure: AI is everywhere, but they’re still not quite sure how to use it. It’s all about the fear of falling behind.
This fear is completely justified. The gap is growing between people who can apply AI in their work and those who can’t. And that gap shows up in speed, opportunities and confidence.
This 2-day AI Mastermind from Outskill is designed to help you start using AI in your day-to-day work (because understanding is never really enough).
What will you get out of it?
Clarity: what tools matter and where to start
Practical skills: writing, automation, basic AI flows
Efficiency: ways to save time in your day-to-day tasks
Career edge: understanding how AI can make you more valuable
You’ll also receive templates, toolkits and guides to help you apply what you learn faster.
Dates: April 11–12 | 10 AM – 7 PM (EST) | Online | Free
Register now
Special highlight from our network
The race to build with AI has produced a surge of output, faster launches, shorter cycles, constant iteration. Yet beneath the pace, a quieter question remains unresolved: are we building what truly matters?
Speed, on its own, is a poor compass.
This session examines a shift taking hold among leading builders: Vibe Solutioning, an approach defined not by how quickly something is shipped, but by how precisely a problem is understood and solved.
Steve Nouri is joined by Vishal Virani (Co-Founder & CEO, Rocket) to discuss the realities of scaling AI products, and the discipline required to turn capability into value.
A conversation for those less interested in momentum, and more concerned with direction.
April 22 · 3 PM CEST | 6:30 PM IST
Anthropic Hits $30B Revenue Run Rate and Overtakes OpenAI
Anthropic's annual revenue run rate has surged past $30B, up from roughly $9B at the end of 2025. That figure now exceeds OpenAI's reported $24-25B. Fewer than 135 S&P 500 companies booked that much in sales over the past 12 months.
Enterprise adoption: The number of business customers spending more than $1M annually has doubled in under two months, from 500 in February to more than 1,000 today.
Compute deal: Anthropic confirmed an expanded partnership with Google and Broadcom, securing 3.5 gigawatts of TPU-based compute capacity from 2027. Most of it will be deployed in the US.
Infrastructure commitment: The deal builds on a $50B American AI infrastructure commitment announced last November. Anthropic runs Claude across AWS Trainium, Google TPUs and NVIDIA GPUs.
Strategic contrast: OpenAI's growth is powered largely by ChatGPT's consumer base. Anthropic has focused on enterprise customers and API usage from day one.
The revenue composition tells different stories about which bets are paying off fastest. OpenAI expects to burn $85B before reaching profitability in 2030. Anthropic projects 2026 as its biggest loss year, with profitability expected by 2029. Both companies are spending at extraordinary scale, but Anthropic just passed its rival while spending less.
Three Fortune 500 CEOs Step Down Citing AI
The CEOs of Walmart, Coca-Cola and Adobe have all announced departures within recent months. All three explicitly cited AI as a factor in their decisions.
Adobe: Shantanu Narayen was the most direct, framing his exit around the next era of creativity being shaped by AI. Adobe's stock was down 23% in 2026 at the time of announcement.
Coca-Cola: James Quincey told CNBC he concluded it was time to put someone else on the field for the next wave of growth, pointing to AI and generative technologies as the transformative shift.
Walmart: Doug McMillon said his successor John Furner was uniquely capable of leading the company through the next AI-driven transformation. McMillon had led Walmart for nearly 12 years.
One reading: convenient timing for seasoned executives cashing out at the top. Another: a signal that the people running three of the largest companies on earth looked at what is coming and decided it needed different leadership. When the CEOs of a retailer, a beverage giant and a software company all point to the same technology in the same breath, something real is happening.
OpenAI Publishes a "Social Contract" for the Intelligence Age
OpenAI released a 13-page policy paper laying out ideas for how society should prepare for superintelligence. It is the most detailed policy document any major AI company has published to date.
Wealth fund: A sovereign-style fund seeded by AI firms that would pay dividends to every American, modelled on Alaska's oil revenue programme.
Robot tax: A proposed tax on robot labour to offset workforce displacement.
Four-day workweek: A formal proposal for reduced working hours as AI productivity gains accelerate.
Right to AI: Universal access to AI tools, framed as a fundamental right.
Containment playbooks: Protocols for managing rogue autonomous systems that cannot be shut off.
The timing is notable. This is the CEO of an $852B company publicly asking the US government to prepare for a future where his own technology breaks the economic system. You do not make that pitch unless you believe it is coming. Whether Washington moves fast enough is another question entirely.
Tool of the Day: Google AI Edge Eloquent
Google has released a free offline dictation app that turns messy speech into clean, polished text. No subscription, no usage cap, no data leaving your phone.
How it works: Runs Gemma-based models on your device, strips filler words like "um" and "ah," corrects mid-sentence stumbles and copies clean text to your clipboard.
Text modes: Transform output into key points, formal, short or long formats with one tap.
Personal dictionary: Import terms from your Gmail or add jargon manually to improve accuracy.
Optional cloud: Toggle on Gemini for enhanced polish, but the core experience is fully offline.
iOS only for now: Android referenced in the listing but not yet live.
Rivals like Wispr Flow and Willow charge $15 a month. This does largely the same thing for free.
Try it yourself: Search "Google AI Edge Eloquent" on the App Store.
Light Bytes
Meta's first Wang-era models: The codenamed "Avocado" model was delayed in March over poor benchmark performance, with some models to be open-sourced and the largest kept closed.
Iran targets Stargate: Iran's military published satellite footage of the $30B Stargate data centre in Abu Dhabi and vowed to destroy it.
AI prescriptions without doctors: Legion Health won approval to let its AI app refill psychiatric medications without clinician oversight, a US first.
Vercel's AI code reviews: 58% of pull requests in Vercel's largest monorepo now merge without human review, cutting average merge time by 62%.
OpenAI's IPO friction: Altman wants to list as soon as Q4 but CFO Sarah Friar reportedly disagreed, and has since been excluded from some financial planning conversations.
OpenAI Image V2: OpenAI is testing its next-gen image model in three variants on ChatGPT and LM Arena, with early improvements in UI rendering and prompt adherence.
