Hegseth declares war on Claude

On Friday, Pete Hegseth, former Fox News personality and current Secretary of Defense, exercised the “nuclear option” and cancelled the Pentagon’s contract with AI startup Anthropic and declared them a “supply chain risk,” effectively barring the company from any federal contracts.

What got the well-coiffed Secretary so wrought up? Anthropic’s AI model, Claude, underpins the data management technology of defense contractor Palantir. US military leaders have embraced Palantir as a way to meld the Pentagon’s fragmented databases into an integrated decision support network for logistics, war planning, intelligence, and more. Anthropic has had a close and productive relationship with Palantir, but stipulated two “red lines” in its contract: 1) no use of Claude’s AI for autonomous killer robots without a human in the loop, and 2) no use for widespread surveillance of the American people. That’s it. That’s the ask - no unguided killer robots, and no mass spying on Americans.

The nattily-groomed Secretary would have none of it. He huffed, he puffed, he threatened to cancel OpenAI’s $200 million per year contract, and to blackball them from federal contracts as a “supply chain risk.” This is a designation typically reserved for technology from adversaries like Russia and China who are suspected of introducing spyware into their products. It is highly unusual to levy this charge against a prominent US company, to say the least. He gave Anthropic two days to knuckle under.

Anthropic didn’t cave. CEO Dario Amodei informed the spokesmodel Secretary that the red lines would hold. Hegseth canceled Anthropic’s contract and designated them a supply chain risk. President Trump directed all federal agencies to cease doing business with Anthropic, giving a 6-month transition period to transfer data and systems to a “more patriotic service.”

Which more patriotic service? Read on, dear reader, the plot twists, then twists again in the next two news items.

 Hegseth declares war on Claude

OpenAI signs Pentagon deal before Claude’s body is cold

On Wednesday, Secretary of Defense Pete Hegseth gave AI startup Anthropic 48 hours to allow the Pentagon to use the company’s Claude chatbot to develop autonomous killer robots, and introduce mass surveillance of the American people.

A bold move, if you have no backup. And on Wednesday, the Pentagon had no contract with a comparably capable AI company. But, guess what? As soon as Hegseth cancelled Anthropic’s contract on Friday, he inked a deal with Anthropic’s arch-rival, OpenAI. What a coincidence! It was almost as if the whole thing was planned in advance.

How could that be? Well, OpenAI CEO Sam Altman is very chummy with Oracle CEO Larry Ellison. Ellison is very chummy with President Trump. (On a completely separate track, Ellison is amassing vast chunks of mainstream news media, including CBS news, CNN, and TikTok under his son’s Paramount brand, in order to give the acquired companies a bent friendly to Trump.)

Somehow, OpenAI was waiting in the wings, and able to sign a deal within hours of Anthropic’s cancellation. Quick work!

OpenAI is desperate for cash, because Altman has been having difficulty monetizing the 800 million monthly users of ChatGPT, 95% of whom use it for free. So a cool $200 million Pentagon contract was very enticing.

Altman tweeted a mealy-mouthed explanation of the deal on X/Twitter, claiming he had put in all the same red lines into his contract with the Pentagon that Anthropic had. Sounds fishy, why would Secretary Hairline agree to that?

And indeed, when more of the actual language of the contract was revealed, it had loopholes you could drive an army of killer robots and mass surveillance drones through.

So Anthropic is out, OpenAI is in, and crony capitalism wins another round? Not so fast - the saga continues.

OpenAI signs Pentagon deal before Claude’s body is cold

Uproar over Pentagon drama - Claude #1 in the App store, OpenAI scrambles, walks back its deal with DoD

As often happens with narcissistic bullies, Hegseth and Trump’s over-the-top manhandling of AI startup Anthropic has engendered a backlash. 

Anthropic immediately filed suit against its designation as a supply chain risk, a designation intended for adversaries like China and Russia, Legal analysts are pretty confident that the designation will not survive contact with the legal system.

The bullying of Anthropic sent shock waves through Silicon Valley, where the callous disregard of law and precedent revealed that any company could be blacklisted at the whim of two posturing man-boys. Silicon Valley tech bros have trended ever more favorable to Trump over the past decade, and this was a stark reminder that the leopards could eat their faces, too. (*This is a reference to a classic political meme; "'I never thought leopards would eat MY face,' sobs woman who voted for the Leopards Eating People's Faces Party.")

And the people began voting with their feet, canceling ChatGPT and signing up for Claude in droves. Claude became the most-downloaded app on the App Store for the first time, outpacing perennial favorite ChatGPT.

Altman scrambled, simultaneously trying to explain away his Pentagon deal, and promising to renegotiate stronger red lines into the contract. It all sounded like tired old damage control and spin.

Backlash over Pentagon drama - Claude #1 on the App Store, OpenAI scrambling

Karpathy says coding changed forever in December 2025

AI legend Andrej Karpathy, former head of AI at Tesla, and a co-founder of OpenAI back in its idealistic early days, recently shared an epiphany he had about coding computer software. Which is, don’t. He now believes that the era of humans hand-coding software is effectively over.

A highly skilled programmer himself, he has closely watched the progress of AI coding assistants over the last 2 years. In December, with the release of Anthropic’s Opus 4.5 and OpenAI’s GPT 5.2, he sensed a sea change. The coding assistants were now so good that software engineers were no longer directly coding at all. Instead, the software engineers spent their time planning what should be coded, setting off multiple AI coding agents on separate coding tasks in parallel, evaluating the functionality of the code produced, and communicating with team members, internal stakeholders, and customers.

In the future, hand-coded software may become a novelty, or a luxury good, like craft beer. And now that computer code is being produced in minutes or hours, rather than days and weeks, AI is improving itself faster and faster, so that model upgrades are becoming almost continuous.

Software engineering is the first profession to be radically transformed by AI. It is unlikely to be the last.

Karpathy says coding changed forever in December 2025

FinTech giant Block lays off 40% of workforce, blames AI

Block, a FinTech powerhouse that invented Square (get it?), the attachment that makes every iPad a cash register for small businesses, just hit record profits for the past quarter. So it’s laying off 4,000 employees, 40% of its entire workforce.

Block CEO Jack Dorsey, a co-founder of Twitter (now X), says AI made him do it. As noted in the story above, AI agents are getting so capable, that a lot of software programming and other routine professional tasks can be done faster and cheaper with AI. The layoff will eventually save Block some $800 million in labor costs, most of which would fall directly to the bottom line, increasing profits by 50+%.

The news and the rosy projections caused Block’s stock to jump 20%. Other company boards and CEOs will get the message. Expect copycat layoffs to proliferate.

Block lays of 40% of workforce on record profits, blames AI

SWARM Biotactics uses cyborg cockroaches for recon

DefenseTech startup SWARM Biotactics has raised a €10 million seed round to develop bio-robotic systems of fully controlled living insects for defense, search and rescue, and intelligence gathering. Each living insect is fitted with an electronic backpack that enables precision control of movement, sensing, and communications. This eliminates the challenging problem of miniaturizing robust robot hardware. The tiny form factor is also optimized by evolution for exploring natural and disturbed environments, and makes the cyborg cockroaches nearly undetectable by conventional means.

SWARM Biotactics uses cyborg cockroaches for recon

China sprints ahead in humanoid robots

China is leveraging its lead in manufacturing technology and supply chains into what may become a commanding lead in humanoid robots. The Chinese government made the decision in its 10th Five-Year Plan (2001- 2005) to pivot from internal combustion engines, where the West was far ahead, and try to “leapfrog” ahead in electric vehicles. The robust Chinese EV manufacturing sector built the supply chain for EVs, from sensors to batteries, and then was able to build on this base to begin experimenting with humanoid robots, iterating faster and at a lower cost than Western rivals could match. In addition, several EV car manufacturers began developing humanoid robots directly, and used them in their automobile assembly plants, gaining invaluable experience while saving on labor costs and giving their robotics division a built-in customer base. The result has been extremely rapid evolution of humanoid robots in China, as demonstrated by the impressive acrobatic feats of robots at the Chinese New Year Gala last month, video clips of which became widespread internet memes.

China sprints ahead in humanoid robots

VC fund A16Z foresees infinite healthcare with AI

Venture capital superfund Andreessen Horowitz - known almost universally as A16z - thinks AI is the cure for what ails healthcare. When each individual can be supported throughout life by a personalized AI, they can be educated to make better choices in diet, exercise, daily habits, and medical care, lowering the incidence and the cost of chronic disease. AI would handle all routine healthcare interactions with individuals, with humans needed only for procedures and face to face interactions. The graph below illustrates their vision - AI raises the total cost of healthcare early on, but the improved health outcomes give compounding returns over time, so that healthcare costs as a percentage of GDP go down. (dotted line.) In this essay, vision is large, practical suggestions for specific changes are few.

VC fund A16z foresees infinite healthcare with AI

Heidi launches Evidence, acquires AutoMedica

Australian ambient AI documentation startup Heidi Health, which has challenged larger HealthTech companies with its grassroots, one doctor at a time initial growth strategy, continues to punch above its weight. It has just launched Evidence, an apparent competitor to OpenEvidence as a clinical decision support tool, which is being integrated into the full Heidi platform rather than being a standalone. Heidi also emphasizes that its offerings are ad-free, so that there is no incentive to bias the content, a clear jab at ad-supported OpenEvidence.

Heidi has also announced the recent acquisition of AutoMedica, a UK-based AI startup that automates clinical and administrative workflows in care. It seems clear that Heidi has ambitions beyond being an AI Scribe, and wants to build what it calls a Care Partner, an AI assistant for every clinician.

Heidi launches Evidence, acquires AutoMedica