19 May 2026: Karpathy Joins Anthropic as the AI Talent Race Enters a New Chapter

Andrej Karpathy, one of the most recognised names in AI research, has started at Anthropic. The same day brought a $300 million acquisition that removes a key tool from rivals, a packed Google I/O keynote, and the end of Elon Musk’s legal battle with OpenAI. Tuesday’s AI news moved at pace.

Andrej Karpathy has joined Anthropic, where he will work on the pre-training team led by Nick Joseph. Karpathy confirmed the move this morning, saying Anthropic will build a new group around him focused on using Claude to accelerate pre-training research. Pre-training is the large-scale process that gives a language model its foundational knowledge before any task-specific tuning begins.

Karpathy was a founding member of OpenAI, left in 2017 to lead Tesla’s Full Self-Driving programme, returned to OpenAI in 2023, then departed again in 2024 to start Eureka Labs, a startup focused on AI in education. The move to Anthropic ends that chapter and brings one of the field’s most cited pre-training researchers into Claude’s foundational development.

The timing is not accidental. Anthropic is closing a fundraising round of at least $30 billion, co-led by Sequoia Capital, Dragoneer, Greenoaks, and Altimeter, at a valuation above $900 billion. That would place it ahead of OpenAI’s most recent valuation of around $852 billion. Hiring Karpathy at the same moment that capital lands signals where Anthropic intends to compete next.

Also announced on Tuesday, Anthropic has acquired Stainless, the SDK-generation startup whose tools have underpinned the developer infrastructure of OpenAI, Google, Cloudflare, and Replicate, for a reported $300 million. Stainless was founded in 2022 by Alex Rattray, a former Stripe engineer. Its software takes an API specification and produces production-ready software development kits across Python, TypeScript, Go, Kotlin, and Java automatically.

Anthropic has confirmed it will wind down Stainless’s hosted products and close access to the SDK generator for new customers. Existing customers keep their generated SDKs and can modify them freely. But rivals who relied on the platform to manage their developer tooling will need to find alternatives. For anyone deciding which AI tools are worth committing to, this consolidation is worth factoring in as Anthropic tightens its developer ecosystem.

Google I/O 2026 opened this morning at Shoreline Amphitheatre in Mountain View, with the main keynote scheduled for 6 PM BST. The event is expected to centre on a major Gemini model update, with reports pointing to either Gemini 3.8 or a full Gemini 4.0. An agentic assistant called Remy is among the more specific leaks, described as a persistent AI capable of completing multi-step tasks independently. If you want context on where AI agents actually stand today, our explainer covers the practical boundaries.

Google has confirmed hardware announcements at the event, including Android XR glasses developed with Samsung, XREAL, Warby Parker, and Gentle Monster. A new laptop platform called Aluminium OS, which merges Android and ChromeOS, is expected to ship on devices from five major manufacturers later this year. Any new Gemini capabilities deployed in Europe will face review under the EU AI Act’s advancing compliance requirements.

A California jury dismissed all of Elon Musk’s claims against OpenAI and CEO Sam Altman on Monday, finding the lawsuit was time-barred by the statute of limitations. The nine-member jury deliberated for less than two hours following eleven days of testimony in Oakland. Judge Yvonne Gonzalez Rogers agreed and dismissed the case. Musk’s primary claims covered breach of charitable trust and unjust enrichment. The jury did not reach the underlying allegations about OpenAI’s mission or governance. Musk called the result a “calendar technicality” and his attorney indicated an appeal is planned.

SandboxAQ, an Alphabet spinout chaired by former Google CEO Eric Schmidt, has integrated its physics-based AI models into Claude via the Model Context Protocol, making advanced drug discovery and molecular simulation tools accessible to researchers without dedicated computing infrastructure. The models are trained on real-world lab data and scientific equations, enabling quantum chemistry calculations and simulation of how drug molecules will behave before lab testing begins.

Before this integration, running a SandboxAQ model required bringing your own compute setup, a barrier that limited use to large pharmaceutical companies and well-resourced labs. Accessing the same tools through Claude now requires only a subscription and a question in plain language. For smaller biotech teams or independent researchers, that is a concrete change in what is practically available to them.

Here is everything else worth knowing from today’s AI news.

• Anthropic closes $30 billion funding round: The raise is co-led by Sequoia, Dragoneer, Greenoaks, and Altimeter at a valuation above $900 billion, reportedly ahead of OpenAI’s most recent figure. Bloomberg
• Andon Labs lets AI run a radio station: The startup launched Andon FM, an experiment in fully autonomous AI-run broadcasting, and is publicly reporting what goes wrong without human oversight. Andon Labs
• Simon Willison recaps six months of LLMs in five minutes: The developer and AI commentator published a dense summary of the most consequential LLM developments since November 2025. Simon Willison
• Five major AI labs signed US government pre-deployment evaluation agreements: Anthropic, OpenAI, Google DeepMind, Microsoft, and xAI finalised agreements with the Commerce Department’s CAISI to allow government testing before frontier model releases. CNN

The clearest thing to watch over the next 24 hours is what Google announces in its Gemini session at I/O today. Karpathy’s move to Anthropic has changed the talent picture publicly. Whether Google responds with a model capability announcement that reframes the competition, or whether today’s keynote plays it conservative with incremental updates and hardware, will tell you a great deal about where each company thinks the real frontier race is currently being decided.

This is a daily news update for informational purposes only. AI products and policies change rapidly. Verify details directly with providers before making decisions. Nothing here is financial or legal advice.

AI Daily is Cristoniq’s afternoon update on developments in artificial intelligence, published every weekday afternoon.