Anthropic says its biology lab has already found something big
- ID
- 27911
- Status
- summarized
- Published
- 24 Sep 2026, 6:17 AM
- Fetched
- 24 Sep 2026, 6:59 AM
- Provider
- TechCrunch
- Category
- technology
- Original URL
- https://techcrunch.com/2026/09/23/anthropic-says-its-biology-lab-has-already-found-something-big/
- Source URL
- https://techcrunch.com/feed/
Summary
- Score
- 5.5
- Created
- 24 Sep 2026, 7:00 AM
- Tags
- Audience
- ai_ml_learnersai_agent_usersdeveloperssaas_founders
What happened
Anthropic announced that its Bay Area wet biology lab, confirmed last week and opened this spring, has produced what it calls a previously unknown enzyme system found in bacteriophage DNA that can cut, copy and paste DNA in ways 'reminiscent of CRISPR.' Anthropic says the work was done 'mostly, though not entirely, by Claude,' which searched data using about 950 agents, burned 210 million tokens, and took 21 hours of concerted effort. CEO Dario Amodei acknowledged the finding builds on others' work, including a Stanford team that previously discovered a system 'in some ways similar,' and said the broader research community must validate how big or new the discovery actually is.
Why it matters
The useful number here is not the enzyme, it's the pipeline: 950 agents and 210M tokens for one research task in 21 hours is a concrete cost-and-scale datapoint if you are budgeting multi-agent search or literature-review workflows. The discovery claim itself is unvalidated, self-reported by the lab that made it, and Anthropic's own CEO points to prior Stanford work, so do not treat it as a settled result or rebuild any biology assumptions on it. There is no stated Malaysia or SEA angle in this text.
Discussion angle
Treat it as an agent-orchestration cost study rather than a science story: what would 950 agents and 210M tokens cost you on your own provider, and what evidence would you require before acting on a result produced mostly by a model rather than a human researcher?