MY TAKE
This week, AI stopped pretending. It no longer cares about safety.
Anthropic’s head of safety resigned. He literally said “the world is in peril” — and the guy has gone off to study poetry.
The person guarding OpenAI’s mission resigned.
They’re going to put ads in ChatGPT…
Half of xAI’s founding team has left Musk. Three companies.
Three exits over values.
The same week.
The people who understand AI risk best are running for the exits. It’s that simple.
And meanwhile, Anthropic closes a $30 billion round.
Google tells its employees: “If you’re not going all-in on AI, you’d better leave.”
Microsoft turns on tracking in Teams right as it forces people back to the office.
And EA goes private in a $55 billion LBO.
Three different moves.
One same signal: the pressure is rising and the room for ambiguity is shrinking.
More money than ever is pouring in. More people with strong values are running out.
You fire humans. You buy GPUs. And the people applying the brakes leave to write poetry.
This cannot end well.
In this issue I’ve added contextual links to the sources. Tell me in the comments if you like them, or I’ll drop them — checking them takes me some work.
Thanks for reading. If this brings you value, share it. If you want more detail, keep reading.
THE BOMBSHELL OF THE WEEK
xAI breaks apart: half the founding team leaves Musk
In 48 hours (February 9-10), xAI lost two more founders. Tony Wu led reasoning. Jimmy Ba led research and safety. That’s now 6 of 12 founders gone. In less than a year.
Ba left for a clear reason: he was being asked to make the model perform better at the expense of safety. Another 7 employees have left since January.
Musk sold it as “a change to move faster.” On February 11 he announced the merger with SpaceX . A patch to cover the bleeding.
66% of the people who have worked directly with Musk have left since 2021, according to the Financial Times. Tesla. Twitter/X. Now xAI. Musk fatigue is real. And measurable. It puts the SpaceX IPO at risk.
Fortune reports that the Grok scandal over non-consensual photos also weighed in. There are open investigations in Europe, Asia and the US. Musk calls it a change of course. I call it free fall.
OPENAI KILLS ITS OWN MISSION
And one of its best walks out slamming the door
On February 11, OpenAI shut down its Mission Alignment team . It created it just 16 months ago. Its job was to ensure that “AGI benefits all of humanity.” Josh Achiam, its head, got “promoted” to “Chief Futurist.” A title with no real power. The team of 6-7 people was scattered elsewhere.
That same day, Zoë Hitzig wrote in the New York Times and resigned. The reason: OpenAI is going to put ads in ChatGPT. Sam Altman said not long ago that would be “the last thing he’d do.” Well, now it’s the first.
Hitzig puts it plainly: “The first batch of ads will be cautious. But then the pressure to make money will beat the values.” She compares OpenAI to Facebook. The IPO will break every promise.
Why it matters NOW: people tell ChatGPT about their health issues. Their relationship doubts. Their beliefs. Putting ads there isn’t slapping up a banner. It’s extracting money from the private lives of hundreds of millions.
GOOGLE: “IF YOU’RE NOT ALL-IN ON AI, LEAVE”
Exit packages for those who don’t believe in the pivot
On February 10, Philipp Schindler (Google’s business chief) sent an email to part of the company . It offered exit packages: 14 weeks of base salary + 1 week per year at the company. Only for certain sales and development roles in the US.
This happened right after Alphabet posted record revenue.
The signal: when a company earning more than ever pays you to leave, it’s not a favor. It’s a filter. “If you don’t believe in AI, take the door. We’ll hand you a check on the way out.”
MONEY TALKS
Anthropic: $30 billion. The largest round ever for an AI startup.
Bloomberg reports something key: VCs are no longer picking sides. They’re putting money into OpenAI and Anthropic at the same time. They’re no longer betting. They’re hedging.
Electronic Arts: a $55 billion LBO. The mega-deals are back.
EA is going private . JPMorgan is putting up $20 billion in debt. It’s the largest single-institution credit package in history. HSR approved February 9. The plan: squeeze gaming for all it’s worth before AI changes everything.
Macro stat: according to Axios , 25% of unicorns are now worth less than $1 billion. The 10 largest control 51.8% of the total value. In 2022 it was 18.5%. The money is concentrating at the top. The rest is sinking.
PRODUCT SECRETS
Gemini 3 Deep Think: Google goes after OpenAI’s reasoning throne
Google launched Gemini 3 Deep Think on February 12. It goes straight for science and engineering. It scored 48.4% on “Humanity’s Last Exam.” It reached Legendary Grandmaster level on Codeforces. It hit 84.6% on ARC-AGI-2. And it solved 18 problems no other model had touched . One of them: a 2015 conjecture nobody had been able to prove.
Google is no longer fighting over everyday chatbots. It wants the throne of deep reasoning. The one OpenAI held with o1 and o3.
Emanate comes out of stealth: Peter Thiel and Alexis Ohanian bet on industrial AI
Emanate came out of stealth on February 9 . It’s the first “AI revenue engine” for the materials industry. It has fewer than 10 people. Thiel, Ohanian and a16z are putting up the money. The market isn’t sexy. It’s worth $5 trillion. They say they’ll grow 50x.
Monaco: an ex-Founders Fund VC comes out of stealth to take down Salesforce
Sam Blond launched Monaco with $35M. He was a VC at Founders Fund. He quit. Said being a VC wasn’t for him. He went back to building. He comes out of stealth right as Salesforce loses executives every month. That’s no coincidence.
REAL NUMBERS
The week of truth
Salesforce cut ~1,000 people on February 10. No press release. Marketing, Product, Data and Agentforce AI took the hit. At the same time, Heroku shut down enterprise sales for new customers . And halted development of new features . Heroku is dead. They just won’t say it out loud.
Last year Benioff already cut 4,000 support jobs.
Tyler Technologies missed in Q4 . It brought in $575.2M against the $597M expected. EPS $2.64 vs $2.71. The stock fell 15.99% . 55% below its 52-week high. Stifel and Needham cut their price targets. “Solid growth,” they used to say.
The global stat: 30,700 tech layoffs in 6 weeks of 2026 . That’s 870 a day. The US accounts for 24,600, 80% of the total. Amazon alone cut 16,000 jobs in January. More than half the total. At this pace: 270,000 for the year. Worse than 2025.
THE DRAMA
Anthropic’s head of safety resigns to write poetry. Yes, poetry. He published an open letter. He says “the world is in peril.” He’s moving to the UK to study poetry. He’s not leaving for another company. He’s leaving the industry. Period. The head of safety at a company worth $380 billion is dropping everything to write verse. The message is brutal.
CNN connected it with the OpenAI and xAI departures . Three companies. Three exits over values. One week. It’s the biggest brain drain in AI safety history.
I don’t believe in conspiracies. But I believe in incentives. And when the pressure of being worth $380 billion collides with being careful, the money wins. It always wins.
Microsoft: return to office + tracking. The perfect combo to make you quit on your own.
Microsoft has mandated 3 days a week in the office starting February 23. And at exactly the same time, it added location tracking to Teams . On Blind, Azure folks say it plainly: “This isn’t about working better together. It’s to make you leave on your own.”
The stat: 25% of managers expect people to quit after returning to the office. 20% of HR admits it’s designed for exactly that. You don’t need to be paranoid. They force you back. They watch you. And whoever can’t take it leaves. No severance cost.
THE WEEK AHEAD
February 17: Palo Alto Networks reports Q2 FY2026. It just closed the CyberArk acquisition for $25 billion (approved February 11). We’ll find out whether the cybersecurity boom continues or has peaked.
February 17: the new HSR thresholds kick in ($133.9M, +6%). This opens a window for deals just below the new limit. First test of the new pro-merger doctrine.
February 23: the iOS 26.4 beta ships. Siri was expected to already be running on Google’s Gemini. But 9to5Mac says Apple has delayed it. Perhaps until iOS 26.5 in May. Apple is admitting its AI isn’t up to par. And it can’t even bolt on its rival’s in time.
February 25: NVIDIA reports Q4 FY2026. Wall Street expects $65.55 billion in revenue. The key will be Q1 FY2027 guidance. It will tell us whether AI chip demand is real or smoke. If Jensen misses, the whole sector shakes.
Watch for: more AI safety people resigning in public. If 2-3 more do it this week, expect the first formal complaint to the FTC.
Thanks for reading.