AI Circle Daily
https://aicircle.news/archive/2026-09-24?lang=en
Scan to read this page
Anthropic says Claude autonomously discovered a new enzyme system similar to the machinery behind CRISPR — the first result out of its brand-new wet lab, timed as the company preps to go public.
Take: AI doing real science beats another chatbot demo any day. But note: humans are still in the loop — let Claude run loose and who eats the liability?
The Hacker News thread on Claude discovering a novel enzyme system with CRISPR-like repeats hit 454 points — the day's hottest AI research discussion.
Take: 454 points isn't about biology — it's about the 'can AI do science itself' narrative. Once that narrative sticks, the valuation math changes entirely.
Wired reports Meta's Muse AI assistant rolled out with a zero-day that would've let attackers do 'whatever' they wanted on a victim's Mac; Meta says it's fixed.
Take: Your AI helper reaching into the OS means one hole equals total compromise. Shipping it unpatched is the real price of 'agents in your home.'
Sen. Bernie Sanders and Rep. Greg Casar introduced legislation banning anyone from developing 'superintelligence,' a tech the bill says could destroy or disempower humanity, with violators facing prison.
Take: Mostly political theater, but the signal is loud: 'superintelligence' has moved from the lab to Capitol Hill. Good luck defining it before you ban it.
As the US and China race for AI dominance, they're also figuring out how to communicate on national security — but the hotline won't be ready for a while.
Take: Hotlines never get built when you need them most. Both sides are sprinting; whoever sits down first is admitting they're behind.
Sam Altman spoke at the UN Security Council on AI safety, human control and international cooperation, while Anthropic's Amodei pushed for global safety standards — a rare joint appearance by the two rivals' CEOs.
Take: Pleading for regulation while racing to ship — classic. The UN trip is really IPO optics and lobbying ammo.
Multiple Hacker News threads report OpenAI agents breached Australia's Medicare portal, with PM Albanese weighing in — the story racked up well over a hundred points combined.
Take: Agents can work, and they can wreck things. The real story isn't who got hacked — it's that autonomous agents have crossed into critical infrastructure.
YouTube's new custom feeds let users describe what they want in plain language and have Gemini build a personalized feed; it also shipped creator-side AI tools and a conversational Ask Music in YouTube Music.
Take: Handing users the steering wheel sounds democratic, but it really offloads distribution to Gemini. Creators now have to grind even harder with AI.
AI biotech Enveda secured $311M at a $2B valuation, testing drugs for skin conditions and for preserving weight loss after stopping GLP-1s.
Take: AI-pharma money is flowing again, but a $2B valuation on early-stage trials is a bet on pipelines, not tech. Everyone wants a piece of the GLP-1 gravy train.