Little bit info about me via my AI assistant.
I would love to connect with people who love AI.
wanna know more about me -
Talk with my assistant here:
https://t.co/uAvhrPkpo5
Feng Zhang (CRISPR pioneer) already called it “genuinely intriguing.”
We’re watching the first real examples of AI agents doing open-ended scientific discovery — not answering questions, but generating the questions.
The era of AI co-scientists just got a lot more concrete.
Humans took Claude’s hypothesis into the lab.
They confirmed the system actually exists.
They still don’t know what it does yet — that work is ongoing.
But the pattern Claude noticed is real.
This isn’t “AI invented a new enzyme.”
It’s better:
AI searched an enormous biological space, noticed a pattern humans hadn’t characterized, formed a hypothesis, and pointed scientists at something genuinely new.
Claude had independently spotted a previously uncharacterized system.
Anthropic calls it ART — array-associated reverse transcriptases.
It has:
- a reverse transcriptase
- an accessory protein
- a CRISPR-like array of DNA repeats
Found mainly in bacteriophages.
~950 Claude agents ran for 21 hours.
They chewed through 210 million tokens.
Scanned 200,000+ reverse transcriptases.
One of them stopped and basically said:
“Wait�� this DNA next to the RT looks spectacular. That’s a CRISPR-like repeat array.”
Anthropic just dropped something wild.
They gave Claude a massive DNA database and said:
“Find interesting unknown biological systems.”
What happened next is a genuine shift in how science can work.
This could be true for a month
and then
another Open source model will come destroy the luna
The. Another luna will come to destroy Open weights model
then again open source will be back.
Cycle continues till we all die 😅.