DeepMind Alumni Startup Claims It Beat Anthropic, OpenAI
A DeepMind-alumni startup says its AI teammate beat Anthropic and OpenAI at replicating research; the claim traces to the company's own framing so far.
What's happening
TechCrunch reports that Inherent, a startup founded by DeepMind alumni, says its AI 'teammate' outperformed Anthropic and OpenAI at replicating research. Separately, Google DeepMind released EmbeddingGemma 2, an open multimodal embedding model, and a Hacker News thread is circulating exit interviews with defectors from OpenAI, Anthropic and DeepMind. The sources tracked include no benchmark results for the startup's claim.
The numbers
No price is recorded for this subject on the story day. Chatter ran at velocity 3.8, with breadth 4 and 6 recent mentions.
Our read
A single-sourced performance claim from the startup itself isn't enough to confirm anything yet. Worth tracking if Inherent publishes benchmarks or if Anthropic or OpenAI respond, but right now this is one company's claim amplified by one write-up.
Context
No price for this subject is recorded on the story day, and the data holds no earlier or later Sort Brick stories on DeepMind to compare with. Chatter ran at velocity 3.8 across 4 threads, with 6 recent mentions. The performance claim comes from Inherent itself and is relayed by one outlet, and the data includes no benchmark results for it. The EmbeddingGemma 2 release and the exit-interview thread are separate items under the same subject. The story carries a 'watch' verdict, and there is no price history to show what followed.
Sources we checked
- Inherent, founded by DeepMind alumni, says its AI ‘teammate’ just outperformed Anthropic and OpenAI at replicating research — techcrunch.com
- EmbeddingGemma 2 is an open multimodal embedding model by Google DeepMind — Hacker News
- Exit interviews with defectors from OpenAI, Anthropic, and DeepMind — Hacker News
See today's live board
This story is frozen as published on October 6, 2026. The live board shows what is moving right now, scored against real money.
Open this story live → See today's live boardMore AI stories
- OpenAI's revenue miss drags the AI trade down October 8, 2026
- Mistral ships a 1-trillion-parameter model October 7, 2026
- Google cuts Nano Banana 2.1 image pricing in half October 7, 2026
- Mistral's 'Le Chonk' Is the Loudest AI Launch Today October 6, 2026
- Altman news this cycle is legal, not business October 5, 2026
- Superintelligence Talk Stays Vague, Thin on Facts October 5, 2026
Research, not financial advice. Sort Brick reports what its data showed at the time of publication. It is not a recommendation to buy or sell anything — you decide.