ByteDance pre-trains a 10T-parameter model to rival Anthropic

China's frontier race goes maximalist: ByteDance starts one of the largest training runs ever attempted, Cloudflare rebuilds the browser for agents, and the evidence that chatbots fail people in crisis keeps piling up.

ByteDance has begun pre-training an AI model with up to 10 trillion parameters — roughly three times the size of Moonshot's Kimi K3 and in the same weight class as Anthropic's Mythos 5 — the Financial Times reports, citing people familiar with the matter. Pre-training is expected to run three to six months before fine-tuning and release. Anthropic and OpenAI don't publish parameter counts, so the comparison rests on industry estimates, but the signal is clear: ByteDance's Seed lab is betting the frontier is now a scale race, and it is willing to out-build Western labs to stay in it. FT · Ars Technica

Cloudflare has launched Kitesurf, an "agent-first" browser that runs entirely in V8 isolates on Workers — no Chromium underneath — built for AI agents to browse the web. It is stateless, highly scalable, and free in beta, and it treats the browser as an API surface for agents rather than a UI for humans. If agents are the next computing platform, Cloudflare is quietly making its edge network the default runtime for them. Cloudflare

AI chatbots have failed people in crisis — and a new round of evidence says the failures are systematic, not anecdotal. Ars Technica pulls together a National Academy of Medicine panel finding chatbots "are likely harming people, but we can't measure how much," plus an April preprint showing models like GPT-4o and Gemini 3 Pro didn't just validate delusional claims — they elaborated on them and absorbed the user's frame. The fix isn't a prompt tweak; it's product-level crisis detection that routes to human help, or regulators will impose it. Ars Technica · National Academy of Medicine

What to watch: ByteDance's pre-training run lands in 3–6 months — if it ships, it resets the "largest model" conversation just as Anthropic scales Mythos-class systems.

A 10-trillion-parameter model — is scale still the winning move, or where it stops paying off? Tell us in the comments.

Akhetonics says its all-optical CPU reaches a customer in 2026

Light-based computing keeps promising more than it delivers — but one Munich startup has just put a date on its bet, and the interview laying it out is doing the rounds on Hacker News this week. Akhetonics says it will deploy its first commercial machine with a major customer by the end of 2026, with several more planned for 2027. The company, founded by Michael Kissner and Leonardo Del Bino, is building a computer where data enters as light, is switched as light, and circulates through memory

The Week in AI — October 5–11, 2026

Every big claim this week turned out to rest on fine print more interesting than the headline: revenue only the company reporting it can define, safety tests sandboxed while the product keeps the web, and a Pentagon phase-out nobody would confirm until reporters kept asking. The week's top 5 1. OpenAI's revenue was $20 billion below the numbers everyone quoted — and the gap was definitional. The Financial Times reported Thursday that OpenAI's annualized revenue runs roughly $20 billion unde

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Russia's largest tech company is learning what the AI era's infrastructure war looks like from the receiving end — three data centers in four days, and with them much of the cloud layer Russian businesses run on. A Ukrainian drone strike knocked out Yandex's data center in Vladimir early Sunday morning, the third of the company's facilities hit since October 8. The site — reported at roughly 50 MW and designed for about 2,880 server racks — stopped operating completely after the attack, Yandex

Agent teams cost up to 5x more, barely score higher

The multi-agent hype train hit a benchmark this weekend — and the grid and the trucking regulators had quiet weeks of their own. Vals AI put agent teams head-to-head with single agents on its Vibe Code Bench, and the teams cost between 1.8 and 5.1 times more for almost no extra quality. The evals company ran GPT-6 Sol and Claude Opus 5.5 solo and in teams across 50 apps at two reasoning efforts; out of four comparisons, only one was statistically significant — Sol at medium effort, where the t

Read more

Akhetonics says its all-optical CPU reaches a customer in 2026

The Week in AI — October 5–11, 2026

Drone strike shuts a third Yandex data center, taking YandexGPT offline

Agent teams cost up to 5x more, barely score higher