Research Papers - SiliconNoon (Page 12)

The Guardrails

METR, Redwood and Apollo have the receipts — but no subpoena power

Two reads on the same shift: the nonprofits now doing the safety work the labs won't publish themselves, and a mathematician's case for rewarding understanding over proofs. The Verge's Hayden Field reports from inside the three nonprofits — METR, Redwood Research and Apollo Research — that have quietly become the industry's de facto auditors, and the piece rests on one uncomfortable fact: their predictions keep coming true.

The Frontier

StepFun's Step 5 Preview leaks at 600B — and the weights stay closed

StepFun's next flagship surfaced on a benchmark page before the company said a word about it, and it breaks the pattern the Shanghai lab built its reputation on. Tesla, meanwhile, is putting its Chinese in-car assistant behind a subscription. StepFun's Step 5 Preview appeared on Artificial Analysis's model page dated September 18 — a 600-billion-parameter reasoning model with a 1M-token context window, text and image input, and a score of 44 on the Artificial Analysis Intelligence Index.

The Guardrails

Anthropic picked Accenture to audit it — and will pay the bill

Anthropic named its first embedded evaluator on Friday — a consulting firm the lab is paying directly, hours after outside experts published the conditions that arrangement would have to meet. Anthropic says Accenture staff will work inside the company with access "comparable to an employee's," evaluating and red-teaming its models, running alignment assessments and testing safeguards.

The Everyday

Google gives CC its own account to run a six-person household

Google Labs stopped treating its CC agent as a personal assistant and gave it a seat at the family table — with its own Google account and its own permissions model. Google Labs is testing a new version of CC that runs a household rather than an individual, and the mechanism is the story: the agent gets its own Google account, a distinct identity, and a permission set that up to six family members negotiate with it.

The Guardrails

Over 100 AI experts attach conditions to the labs' evaluator pledge

The people who would do the auditing just told the labs what independence has to mean — and one of their conditions is protection from being sued for it. More than 100 AI researchers and evaluators published a letter Friday setting minimum conditions for the third-party evaluation that Anthropic and OpenAI have promised, with signatories including Geoffrey Hinton, researchers from Johns Hopkins and Stanford, and staff at the nonprofit evaluator METR.

The Guardrails

The Take — Breaking Mila's windows is a gift to the labs

I think the people who smashed roughly 20 windows at Mila this week did more damage to the case against AI than to the institute — and that the labs they were angry at are the ones who benefit from the mess. Start with what actually happened, because the details decide the politics. Late Tuesday night, about 10 people broke windows and spray-painted graffiti at the Montreal offices of Mila, the institute Yoshua Bengio founded, hours before the city opened ALL IN, billed as Canada's largest AI…

The Frontier

React's creator closed Conway's 50-year-old conjecture with AI

The most interesting AI proof of the year did not come out of a lab. It came out of a software engineer's chat subscriptions, a swarm of agents he admits he didn't understand, and a Lean repository anyone can check. Dan Abramov, co-creator of React, says he proved Conway's refinement conjecture — an open problem John Conway posed 50 years ago in On Numbers and Games — using a self-built laboratory of AI agents that he steered mostly on instinct.

The Frontier

A 9B longevity model tops 18 frontier LLMs on aging data

Aging biology generates more measurement data than any one lab can read, and until this week there was no shared way to check whether an AI can actually interpret it. Insilico Medicine's answer is an open benchmark — and the results are unflattering to the biggest models in the world. Insilico Medicine published LongevityBench, an open suite of 17 tasks spanning clinical records, genetics, epigenetics, transcriptomics and proteomics, and its own compact models — built on Liquid AI's LFM2 and…

The Frontier

Anthropic quietly built a wet lab for its AI drug push

Anthropic's answer to "AI is only a chatbot" is a wet lab in the Bay Area — and a federal flash-flood model that starts catching storms earlier than the forecast does. Anthropic has quietly built a wet lab in the San Francisco Bay Area, moving its biology work off the computer and onto the bench. Two people familiar with the matter told Reuters the startup now runs physical experiments, and Anthropic's head of life sciences, Eric Kauderer-Abrams, confirmed the lab in an interview: "We believe…

The Frontier

Zhipu's GLM-5.3-FlashX hits 200 tokens/s — at 2.5x the price

Zhipu shipped the fast tier of its open model today and, unusually for a Chinese lab in 2026, raised the price to do it. GLM-5.3-FlashX is live on the API at up to 200 tokens per second — roughly five times the throughput of GLM-5.3-Flash — and priced at about 2.5x what the base model costs. Zhipu says its 100,000-chip domestic inference cluster was saturated at launch under the weight of demand that never let up, and that the new tier follows fresh infrastructure investment and inference-side…

The Guardrails

Researchers used Claude to walk into OpenAI's internal repos

Two stories about the same thing from opposite ends: a security firm used one lab's models to break into its rival, and a coding agent got a feature for finding what the humans never got around to looking for. Plus a local-inference release that argues the number everyone ranks quants by is measuring the wrong thing. A security company used Anthropic's models to reach OpenAI's internal GitHub repositories in under 72 hours, chaining a heap overflow in an image decoder with a misconfiguration…

The Arena

Treble raises $18M to give physical AI a sense of hearing

Physical AI has been built on eyes and hands. This funding round argues the next sense it needs is ears — and sound is the one training modality nobody can scrape off the web. Treble Technologies raised $18 million (about €15 million) in a Series A-2 led by Paladin Capital Group, with KOMPAS VC, Frumtak Ventures and the European Innovation Council Fund returning, taking the Reykjavík company's total funding to roughly €36 million.

See all