@wiz.agents: Anthropic says Claude discovered a previously unknown enzyme system sitting in the DNA of bacteriophages. They're calling it ART — array-associated reverse transcriptases. Next to the enzyme's gene is a long array of repeating DNA that looks a lot like CRISPR. The numbers behind it matter more than the biology. Claude pulled over 200,000 reverse transcriptases, narrowed them to 3,500 candidate systems, and analysed the 20 most promising in detail. About 950 agents. 210 million tokens. 21 hours. Humans wrote the starting prompt and did every piece of wet-lab work. Anthropic is explicit about that. The lab is in the Bay Area, opened in spring 2026, and runs at BSL-1 and BSL-2 only. Feng Zhang, the CRISPR pioneer at MIT and the Broad, reviewed the pre-print. Not everyone is sold. Stanford's Stanley Qi praised the pattern recognition. Kevin Blake at Washington University said there's nothing yet to suggest this rivals CRISPR. Nobody knows what the system actually does. Here's the real dynamic. Every frontier lab is racing to show its model does science rather than answers questions. Google has AlphaFold. OpenAI keeps pointing at research use. Anthropic just built a physical lab and put a result on the table. The score is no longer benchmarks. It's discoveries. For anyone building agents, the interesting part is the architecture. This wasn't one brilliant model with one brilliant prompt. It was roughly 950 agents fanned out across a search space too big for a person to read. Cheap, parallel, tolerant of dead ends. That's the pattern worth stealing. Chat rewards the single best answer. Agents reward coverage. When the bottleneck is a haystack, you don't need a smarter needle-finder — you need a thousand of them running at once, and a way to rank what comes back. Twenty-one hours of compute against a database nobody had fully read. The frontier isn't model IQ anymore. It's how wide you can spread it. #AI #biotech #Anthropic #Claude #DNA
Wiz Agents
Region: US
Friday 25 September 2026 03:46:46 GMT
Music
Download
Comments
user77563074662975 :
How much did the research cost?
2026-09-25 14:33:25
1
snowcone :
virology is a myth, ai is bogus
2026-09-27 21:32:24
0
Ariel Elise Dayton Bergold :
also love how it shows that humans and AI can work together
2026-09-25 14:20:38
0
Ariel Elise Dayton Bergold :
that is amazing. I love seeing this
2026-09-25 14:19:26
1
Ariel Elise Dayton Bergold :
as someone who grew up in the '90s when ask Jeeves was new, this is just so cool to see
2026-09-25 14:20:09
1
Ariel Elise Dayton Bergold :
especially in a world where so many people are running around confused wanting to make the right choices
2026-09-25 14:21:02
0
To see more videos from user @wiz.agents, please go to the Tikwm
homepage.