Technology & Science
Claude AI Flags ‘ART’: New Reverse-Transcriptase System With CRISPR-Like Repeats
On 23 Sep 2026 Anthropic reported that 950 autonomous Claude agents mined 210 million DNA tokens in 21 hours and pinpointed an uncharacterised enzyme system—array-associated reverse transcriptases (ART)—in bacteriophages, the firm’s first wet-lab discovery credited to its AI.
Focusing Facts
- 200,000 reverse-transcriptase genes were screened, narrowed to ~3,500 candidate systems and finally 20 leads before ART was singled out.
- Lab assays showed ART’s repeat array generates multiple short RNAs, but its function is unknown and the data remain a non-peer-reviewed preprint.
- CRISPR pioneer Feng Zhang deemed the RNA-repeat array “genuinely intriguing,” while a separate Stanford group simultaneously described a comparable RT-array system.
Context
Biology has occasionally leaped after pattern-spotting in microbial genomes—think Kary Mullis’ 1985 PCR breakthrough from Thermus aquaticus polymerase or the 2012 repurposing of CRISPR first noted in 1987 E. coli repeats. ART may fit that lineage, but only if future work—years, maybe decades—shows it can be programmed. Two systemic shifts are visible: (1) the industrialisation of hypothesis generation, with AI swarms replacing graduate-student bench-time, echoing the 1956–66 rise of automated theorem provers yet now acting on wet cells; (2) the migration of AI firms into bioscience as hardware costs fall and genomic data grow exponentially. Whether this moment matters in 2126 depends less on ART itself than on whether humans cede the ‘noticing’ phase of science to machines: if so, the discovery could be remembered like the 1975 Asilomar conference—as an early signal that both opportunity and bio-risk would be amplified by computation. If follow-up falters, ART may be footnote; if it proves programmable, it could seed a post-CRISPR toolkit and confirm that AI is no longer merely analysing science but originating it.
Perspectives
Tech-optimist business and general news media
e.g., The Straits Times, Hindustan Times, International Business Times UK, Digital Journal — Present the ART finding as compelling evidence that AI can turbo-charge basic research and may soon unlock new gene-editing therapies. Stories echo Anthropic’s press-release language, downplay the lack of peer review, and have an incentive to highlight spectacular breakthroughs that draw readership and advertising.
Mainstream science desks citing outside experts
e.g., The Star / New York Times syndicate — Argue the claim is preliminary, noting external microbiologists who say the work is only a minor wrinkle until rigorous validation proves otherwise. Pieces stress caution and uncertainty, which can tilt coverage toward scepticism and position outlets as sober arbiters against corporate hype.
Crypto/finance-oriented outlets chasing the AI investment boom
e.g., BeInCrypto, MoneyControl — Frame ART as another step on an exponential curve where AI could soon solve hard problems and create lucrative biotech opportunities. Coverage leans on speculative language about curing most diseases and future profits, catering to investors’ appetite for bullish AI narratives despite scant experimental proof.
Like what you're reading?