Sam Altman
ASML
Pentagon
Anthropic dropped a new flagship model, shipped a design tool, and watched Figma's stock get taken out back. The company that built its brand on privacy now wants your passport and a live selfie. Sam Altman finally broke silence on the firebomb attack. He pointed the finger at Anthropic. And OpenAI shipped a model built for biology that could cut drug discovery timelines in half. Let's get into it.
🚀 Anthropic Had the Biggest Product Week in AI History. Figma Paid the Bill.
Three launches in four days. A board resignation three days before the competing product dropped. A 7% one-day crater in an $18 billion public company. If you want to know what "moving up the stack" actually looks like, this was the tape.
April 14: Anthropic Chief Product Officer Mike Krieger resigned from Figma's board, disclosed to the SEC the same day The Information reported Anthropic was shipping a design tool. Fiduciary duty forcing the timing. Everyone understood what was coming.
April 16: Claude Opus 4.7 launched. New xhigh effort level. 1M context at standard pricing. Vision resolution jumped 3x to 2,576px. SWE-bench 87.6%. Cyber safeguards built in as Anthropic tests guardrails it plans to eventually extend to Mythos-class models. CNBC coverage →
April 17: Claude Design shipped under the new Anthropic Labs sub-brand. Text prompts in, polished prototypes, slide decks, one-pagers, UI mockups out. Reads your codebase, builds a design system, exports to Canva, PDF, PPTX, or hands off directly to Claude Code. Datadog says it compressed week-long design cycles into single conversations.
Figma stock fell 7% that day. Adobe dropped too. Wix and GoDaddy were already down 3-4% on the news. VentureBeat framed it plainly: Anthropic has moved from model provider to full-stack product company.
The kicker: Anthropic reportedly turned away unsolicited investor offers this week valuing the company at around $800 billion. More than double the $380B from February.
April 14: Anthropic Chief Product Officer Mike Krieger resigned from Figma's board, disclosed to the SEC the same day The Information reported Anthropic was shipping a design tool. Fiduciary duty forcing the timing. Everyone understood what was coming.
April 16: Claude Opus 4.7 launched. New xhigh effort level. 1M context at standard pricing. Vision resolution jumped 3x to 2,576px. SWE-bench 87.6%. Cyber safeguards built in as Anthropic tests guardrails it plans to eventually extend to Mythos-class models. CNBC coverage →
April 17: Claude Design shipped under the new Anthropic Labs sub-brand. Text prompts in, polished prototypes, slide decks, one-pagers, UI mockups out. Reads your codebase, builds a design system, exports to Canva, PDF, PPTX, or hands off directly to Claude Code. Datadog says it compressed week-long design cycles into single conversations.
Figma stock fell 7% that day. Adobe dropped too. Wix and GoDaddy were already down 3-4% on the news. VentureBeat framed it plainly: Anthropic has moved from model provider to full-stack product company.
The kicker: Anthropic reportedly turned away unsolicited investor offers this week valuing the company at around $800 billion. More than double the $380B from February.
My take:
The Krieger resignation is the detail I keep coming back to. Three days' notice before a product launch isn't a coincidence. It's the legal tell that says "we knew this was coming and we let the market figure it out." Claude Design isn't the best design tool yet. The editing has rough edges, collaboration is basic, there's no GA date. But Anthropic doesn't need to beat Figma at Figma's game. They need to eat the 80% of design work that never makes it into Figma because product managers, founders, and marketers can't use it. That market is much bigger, and it's the one Anthropic is actually going for.
📇 You Switched to Claude for Privacy. Now It Wants Your Passport.
On April 14, Anthropic published identity verification requirements for Claude. A government-issued photo ID and, in some cases, a live selfie. Processed by Persona Identities, a KYC firm backed by Peter Thiel's Founders Fund.
Claude is now the first major AI chatbot to require this. ChatGPT and Gemini do not. The checks aren't universal yet. Anthropic says they trigger on four conditions: repeated usage policy violations, access from unsupported locations (China, Russia, Iran, North Korea, Belarus), terms of service breaches, and suspected under-18 users.
The timing stings. Two months ago, millions of users fled ChatGPT for Claude specifically because Anthropic refused the Pentagon mass-surveillance deal. Free signups jumped 60% since January. Now those privacy-conscious users are being asked to send a passport selfie to a Thiel-backed vendor. Decrypt flagged the irony in a headline: "You Switched to Claude Over Surveillance Fears. Now It Wants Your Passport."
Chinese users without passports are effectively locked out. China-issued national ID cards aren't on the accepted list. The South China Morning Post reports a black market for workaround services has scaled fast in the days since.
Anthropic says the data stays with Persona, isn't used to train models, and isn't shared for marketing. But that's cold comfort to users remembering the October 2025 Discord breach that exposed 70,000 government IDs through the exact same vendor class.
Claude is now the first major AI chatbot to require this. ChatGPT and Gemini do not. The checks aren't universal yet. Anthropic says they trigger on four conditions: repeated usage policy violations, access from unsupported locations (China, Russia, Iran, North Korea, Belarus), terms of service breaches, and suspected under-18 users.
The timing stings. Two months ago, millions of users fled ChatGPT for Claude specifically because Anthropic refused the Pentagon mass-surveillance deal. Free signups jumped 60% since January. Now those privacy-conscious users are being asked to send a passport selfie to a Thiel-backed vendor. Decrypt flagged the irony in a headline: "You Switched to Claude Over Surveillance Fears. Now It Wants Your Passport."
Chinese users without passports are effectively locked out. China-issued national ID cards aren't on the accepted list. The South China Morning Post reports a black market for workaround services has scaled fast in the days since.
Anthropic says the data stays with Persona, isn't used to train models, and isn't shared for marketing. But that's cold comfort to users remembering the October 2025 Discord breach that exposed 70,000 government IDs through the exact same vendor class.
My take:
This is a real tradeoff. Preventing abuse, blocking minors, enforcing export restrictions. All legitimate. But Anthropic's entire brand positioning has been "the AI company that respects you." Asking the people who chose you for that reason to submit biometrics and government ID to a third party is going to cost some trust no matter how cleanly the data is handled. OpenAI's competitive team is sending this article around the office right now.
📢 Altman Breaks Silence on the Firebomb Attack. Names Anthropic.
Two weeks after a 20-year-old threw a Molotov cocktail at his San Francisco home and another car tried to shoot it up 48 hours later, Sam Altman went on Ashlee Vance's Core Memory podcast this week and finally talked about it.
He pointed at doomerism. He pointed at the broader AI discourse. And then he pointed at Anthropic directly: "I think the way Anthropic talks about OpenAI doesn't help."
On Mythos, he went harder. He called Anthropic's decision to withhold the model "fear-based marketing." His framing: the "we built something too dangerous to release, so buy our $100M enterprise package" pitch is a sales tactic dressed up as safety. Anthropic has not responded.
The rivalry has been building for months. December: Amodei mocking labs that declare "code reds." February: Anthropic's Super Bowl ad campaign roasting ads in AI chatbots, landing the same week OpenAI was preparing to put ads in ChatGPT. That month in New Delhi: Altman and Amodei visibly refused to shake hands at an AI conference. In between: Altman saying Anthropic "serves an expensive product to rich people," and Amodei reportedly writing in an internal memo that OpenAI's safety messaging is "safety theatre."
Now a kill list with multiple AI CEO addresses is circulating in federal court filings, and the two biggest AI companies are publicly blaming each other's rhetoric for the environment that produced it.
He pointed at doomerism. He pointed at the broader AI discourse. And then he pointed at Anthropic directly: "I think the way Anthropic talks about OpenAI doesn't help."
On Mythos, he went harder. He called Anthropic's decision to withhold the model "fear-based marketing." His framing: the "we built something too dangerous to release, so buy our $100M enterprise package" pitch is a sales tactic dressed up as safety. Anthropic has not responded.
The rivalry has been building for months. December: Amodei mocking labs that declare "code reds." February: Anthropic's Super Bowl ad campaign roasting ads in AI chatbots, landing the same week OpenAI was preparing to put ads in ChatGPT. That month in New Delhi: Altman and Amodei visibly refused to shake hands at an AI conference. In between: Altman saying Anthropic "serves an expensive product to rich people," and Amodei reportedly writing in an internal memo that OpenAI's safety messaging is "safety theatre."
Now a kill list with multiple AI CEO addresses is circulating in federal court filings, and the two biggest AI companies are publicly blaming each other's rhetoric for the environment that produced it.
My take:
Altman has a point about Mythos marketing and a point about rhetoric. Both can be true. But pointing at Anthropic in the same breath as talking about the guy who tried to burn your house down is a choice. This moment needed something unifying. Instead we got "I think you contributed to this." Two weeks on from the attack and the gap between what AI companies say publicly and what they say about each other is as wide as ever. That's the environment radicalization grows in.
🧬 OpenAI Shipped a Model Built for Biology. Pharma Is Already Piloting.
While Anthropic was owning headlines, OpenAI shipped two products worth your attention.
GPT-Rosalind (April 16) is a frontier reasoning model built for biology, drug discovery, and translational medicine. It's tuned for chemistry, protein engineering, and genomics. These are the scientific workflows where target-to-approval timelines currently run 10 to 15 years. Amgen and a roster of pharma and biotech partners are already piloting it. OpenAI's release notes →
ChatGPT Images 2.0 (April 21) launched what OpenAI is calling its first image model with native "thinking." It reasons through complex visual tasks, verifies its own outputs, and generates up to eight coherent images from a single prompt. 2K resolution, multi-image consistency, available via API as gpt-image-2. DALL-E 2 and 3 are being retired May 12. The New Stack has the full breakdown.
Context: OpenAI closed its $122B funding round at an $852B post-money valuation, revenue is now running at $2B/month ($24B+ annualized), and the ChatGPT ads pilot crossed $100M ARR in under six weeks. IPO is targeting Q4 2026 at roughly $1 trillion. Anthropic may be winning the news cycle, but OpenAI is still the biggest company in AI by almost every measure that matters except narrative.
GPT-Rosalind (April 16) is a frontier reasoning model built for biology, drug discovery, and translational medicine. It's tuned for chemistry, protein engineering, and genomics. These are the scientific workflows where target-to-approval timelines currently run 10 to 15 years. Amgen and a roster of pharma and biotech partners are already piloting it. OpenAI's release notes →
ChatGPT Images 2.0 (April 21) launched what OpenAI is calling its first image model with native "thinking." It reasons through complex visual tasks, verifies its own outputs, and generates up to eight coherent images from a single prompt. 2K resolution, multi-image consistency, available via API as gpt-image-2. DALL-E 2 and 3 are being retired May 12. The New Stack has the full breakdown.
Context: OpenAI closed its $122B funding round at an $852B post-money valuation, revenue is now running at $2B/month ($24B+ annualized), and the ChatGPT ads pilot crossed $100M ARR in under six weeks. IPO is targeting Q4 2026 at roughly $1 trillion. Anthropic may be winning the news cycle, but OpenAI is still the biggest company in AI by almost every measure that matters except narrative.
⚡ Quick Hits
NC Treasurer's office goes all-in on AI, reportedly negotiating to buy Claude. Brad Briner expanded AI deployment across the entire department this week after a ChatGPT pilot showed 10% productivity gains. An information sheet from his office confirmed the NC Investment Authority is separately negotiating a purchase of Claude from Anthropic. State employees' union raised concerns about job impact. Route Fifty →
Claude Mythos goes to the NSA. Axios reported this week the National Security Agency is now using Mythos, weeks after the Pentagon blacklisted Anthropic. Dario Amodei visited the White House last week for what both sides called a "productive" meeting. Foreign Policy →
Corning and Meta broke ground earlier this month on a $267M AI fiber-optic expansion in Hickory. Part of a $6B agreement. 132 new jobs averaging $65K. Hickory is now home to the largest fiber-optic plant in the world, built specifically to serve AI data center demand. Business NC →
Tech layoffs from AI hit 45,000 nationwide in 2026 so far. Challenger, Gray & Christmas said at least half are directly attributable to AI. 47% of college students have considered changing their major because of AI. 16% already have. WRAL →
Hitachi Energy announced a $10M Power Electronics Center of Competence in Cary on April 2, creating 150 new jobs. Purpose: meet AI data center electricity demand, which is projected to grow 35-50% by 2040. Press release →
Google shipped six AI product updates this week. Flow Music (renamed from ProducerAI) powered by Lyria 3 Pro. Free NEET medical school mock exams inside Gemini for 2.3M Indian students. None of them tier-one on their own. Together, they show the most coordinated AI product strategy of any lab right now. Signal vs noise breakdown →
Claude Mythos goes to the NSA. Axios reported this week the National Security Agency is now using Mythos, weeks after the Pentagon blacklisted Anthropic. Dario Amodei visited the White House last week for what both sides called a "productive" meeting. Foreign Policy →
Corning and Meta broke ground earlier this month on a $267M AI fiber-optic expansion in Hickory. Part of a $6B agreement. 132 new jobs averaging $65K. Hickory is now home to the largest fiber-optic plant in the world, built specifically to serve AI data center demand. Business NC →
Tech layoffs from AI hit 45,000 nationwide in 2026 so far. Challenger, Gray & Christmas said at least half are directly attributable to AI. 47% of college students have considered changing their major because of AI. 16% already have. WRAL →
Hitachi Energy announced a $10M Power Electronics Center of Competence in Cary on April 2, creating 150 new jobs. Purpose: meet AI data center electricity demand, which is projected to grow 35-50% by 2040. Press release →
Google shipped six AI product updates this week. Flow Music (renamed from ProducerAI) powered by Lyria 3 Pro. Free NEET medical school mock exams inside Gemini for 2.3M Indian students. None of them tier-one on their own. Together, they show the most coordinated AI product strategy of any lab right now. Signal vs noise breakdown →
🧠 One Thing I'm Thinking About
Look at the cadence. Project Glasswing and Mythos Preview on April 7. Identity verification requirements on April 14. Opus 4.7 on April 16. Claude Design on April 17. Four major launches in 10 days.
That's not a research lab. That's a product company.
When OpenAI shipped that fast in 2023, they were called reckless. When Google ships that fast now, they're called desperate. When Anthropic does it, it's called inevitable.
The question I can't stop thinking about: what happens to a safety-first AI lab when it's shipping four products a month? The brand that built Anthropic was "we take our time, we do it right, we say no to things like the Pentagon deal." The company shipping right now is on a release schedule that doesn't leave much room for the long, patient work that brand was built on.
Maybe you can do both. Maybe shipping fast and shipping safely aren't as incompatible as they seemed. But the brand contract with users who came over from OpenAI for "the careful one" is going to get tested. The passport requirement is one test. Claude Design was another. There will be more. How Anthropic handles the next one tells you whether the product company is still the safety company.
That's not a research lab. That's a product company.
When OpenAI shipped that fast in 2023, they were called reckless. When Google ships that fast now, they're called desperate. When Anthropic does it, it's called inevitable.
The question I can't stop thinking about: what happens to a safety-first AI lab when it's shipping four products a month? The brand that built Anthropic was "we take our time, we do it right, we say no to things like the Pentagon deal." The company shipping right now is on a release schedule that doesn't leave much room for the long, patient work that brand was built on.
Maybe you can do both. Maybe shipping fast and shipping safely aren't as incompatible as they seemed. But the brand contract with users who came over from OpenAI for "the careful one" is going to get tested. The passport requirement is one test. Claude Design was another. There will be more. How Anthropic handles the next one tells you whether the product company is still the safety company.
📅 What's Coming
April 30: StrictlyVC SF, first of 2026
May 12: DALL-E 2 and 3 formally retired
June-July: FIFA World Cup across North America
July 17: Claude Design onboarding credits expire
Q4 2026: Anthropic IPO (reported target)
Q4 2026: OpenAI IPO (reported $1T target)
H2 2026: NVIDIA Vera Rubin platform ships
H2 2026: OpenAI "Sweetpea" consumer hardware unveil
May 12: DALL-E 2 and 3 formally retired
June-July: FIFA World Cup across North America
July 17: Claude Design onboarding credits expire
Q4 2026: Anthropic IPO (reported target)
Q4 2026: OpenAI IPO (reported $1T target)
H2 2026: NVIDIA Vera Rubin platform ships
H2 2026: OpenAI "Sweetpea" consumer hardware unveil
📍 Meanwhile in the Triangle
UNC-Chapel Hill hosted its inaugural AI for Public Good Conference on April 13, bringing together OpenAI's chief economist Ronnie Chatterji (who also teaches at Fuqua), State Treasurer Brad Briner, and researchers across the UNC system. NC Newsline has the recap. The most telling moment: Chatterji admitted he doesn't know what to tell the Class of 2026 at his upcoming graduation speech. "If you graduate in a time of great technology disruption, the legacy of that will affect your job 15 years from now."
Meanwhile, Duke's AI strategic plan is moving from document to deployment. Nearly two years of faculty work went into it. Now comes shared infrastructure, new governance structures, and faculty hires. The state Treasurer's office is in active negotiations to purchase Claude from Anthropic. Corning and Meta broke ground earlier this month on AI fiber infrastructure in Hickory that makes NC one of the most important states in the AI supply chain.
Two months ago, "AI hub" felt aspirational for the Triangle. Now it feels descriptive.
Forward this to someone building with AI in the Triangle. Or subscribe at bullcity.ai →
Meanwhile, Duke's AI strategic plan is moving from document to deployment. Nearly two years of faculty work went into it. Now comes shared infrastructure, new governance structures, and faculty hires. The state Treasurer's office is in active negotiations to purchase Claude from Anthropic. Corning and Meta broke ground earlier this month on AI fiber infrastructure in Hickory that makes NC one of the most important states in the AI supply chain.
Two months ago, "AI hub" felt aspirational for the Triangle. Now it feels descriptive.
Forward this to someone building with AI in the Triangle. Or subscribe at bullcity.ai →