CoCo Newsletter · Issue 21

What’s actually happening in AI, and who is building it.

A read on the people, products, and capital shaping the AI industry. Real names, real numbers, and what they mean for anyone shipping today.

Founders

Mira Murati’s Thinking Machines Lab raised about $2B at a $12B valuation, the largest seed round on record.

Murati, who served as OpenAI’s CTO for nearly seven years before resigning in September 2024, announced Thinking Machines Lab in February 2025. The company confirmed a roughly $2B seed round in July 2025, reportedly led by Andreessen Horowitz at a valuation around $10 to $12 billion.

The lab has assembled a roster of senior ex-OpenAI researchers including John Schulman (who himself left OpenAI for Anthropic in 2024 before joining Murati), Barret Zoph, and Bob McGrew, alongside others. Their stated focus is on building more capable, customizable, and broadly useful AI systems with a research-first orientation. The round is among the largest pre-product seed rounds in technology history.

Reported by Reuters and The Information · Round closed July 2025
Latest in AI
Models

DeepSeek R1 reset expectations on reasoning models, and on training costs.

The Chinese AI lab released R1 in January 2025 as an open-weight reasoning model with performance competitive with OpenAI’s o1, reportedly trained at a fraction of frontier-model costs. The release triggered a public market reassessment, with Nvidia briefly losing more than $500B in market cap on January 27, 2025.

DeepSeek · January 2025
Models

Anthropic ships Claude 4 family, with extended thinking and stronger coding.

Released in May 2025, the Claude 4 generation introduced extended thinking time as a controllable parameter and pushed the bar on agentic coding tasks. Sonnet 4 became a default for many engineering workflows; Opus 4 was positioned for the hardest reasoning and research workloads.

Anthropic · May 2025
Models

Google’s Gemini 2.5 Pro pushes long context into the 1 million token range.

Released in March 2025, Gemini 2.5 Pro became one of the first generally available frontier models to operate reliably at the 1M token mark. The size unlocked workflows like full-repository code review, multi-document legal research, and long video analysis without external retrieval.

Google DeepMind · March 2025
Open Weights

Meta’s Llama 4 brings native mixture-of-experts to open-weights.

Released in April 2025, Llama 4 was Meta’s first major generation built on a mixture-of-experts architecture from the ground up. The release covered Maverick (smaller, edge-deployable) and Behemoth (frontier scale), continuing Meta’s push to keep frontier capability open and inspectable.

Meta AI · April 2025
Founders & people we’re watching
Recognition

Demis Hassabis and John Jumper share the 2024 Nobel Prize in Chemistry for AlphaFold.

Awarded in October 2024, jointly with David Baker, the prize recognized AlphaFold’s impact on predicting protein structures, accelerating drug discovery and biomedical research. Hassabis is co-founder and CEO of Google DeepMind; Jumper led the AlphaFold team.

Nobel Foundation · October 2024
Recognition

Geoffrey Hinton wins the 2024 Nobel Prize in Physics with John Hopfield.

Awarded in October 2024 for foundational work on neural networks and machine learning. Hinton, often called a godfather of deep learning, left Google in 2023 to speak more openly about AI safety concerns. The prize cemented neural networks as a recognized scientific breakthrough.

Nobel Foundation · October 2024
Founder

Ilya Sutskever’s Safe Superintelligence reportedly valued above $30B in 2025.

Sutskever, an OpenAI co-founder, left in May 2024 and started Safe Superintelligence with Daniel Gross and Daniel Levy. SSI raised $1B at a $5B valuation in September 2024. Reports in 2025 placed a subsequent round at a valuation above $30B, all without a publicly announced product.

Reported by Reuters and Financial Times · 2024 to 2025
Essay

Dario Amodei’s “Machines of Loving Grace” lays out an optimistic case for advanced AI.

Published in October 2024, the Anthropic CEO’s essay argues that powerful AI, if developed responsibly, could compress decades of progress in biology, neuroscience, mental health, and economic development into a few years. It became one of the most-cited industry essays of the year.

darioamodei.com · October 2024
Companies & tools shipping

Get this in your inbox.

One email a week. Startups forming, AI shifts that matter, and the people building. No fluff.