Parameter Update: 2026-02

Even with CES, this was a bit of a slow news week.

CES 2026

While last weeek's CES brought a slew of mostly useless AI-gadgets (and talks of a potential solid-state battery breakthrough), we also got some news about Boston Dynamics' new Atlas robot (including DeepMind as a major customer and a planned rollout in Hyundai’s factories in 2027)

Anthropic: Claude Code crackdown

In less exciting news, Anthropic has started cracking down on external tools using the Claude Code authentication for free inference (for people with Pro subscriptions). This seemed inevitable, but also means a lot of hobbyist projects breaking. At least in my Twitter timeline, it seems to have cost them a surprising amount of goodwill. In at least one instance, it has also resulted in a project switching over to officially endorsed OpenAI Oauth instead, which I can't believe was what Anthropic intended.

OpenAI: ChatGPT Health

While people are still debating the utility of LLMs in the medical domain, OpenAI has now gone ahead and launched "ChatGPT Health" - a dedicated space for health questions in ChatGPT. It's limited to a waitlist for users "outside the European Economic Area, Switzerland, and the United Kingdom" for now, but the fact that it will happily sync with Apple Health and other integrations makes me think that OpenAI might have pretty major ambitions here. Personally, I made heavy use of ChatGPT when I got sick over the holidays, so this seems like it might end up being really useful.