Writing
Following the Navier-Stokes controversy, a primer on the privacy and data-retention policies of OpenAI, Anthropic, Google, and xAI: the training defaults by tier, the feedback and safety-flag carve-outs, and what deleting a chat actually does.
AI Safety: What's Real, What's Likely, and What We Can Do
My thoughts after going on BBC to talk about the Hugging Face incident and Jacob Coxon's resignation from Anthropic: the three kinds of danger, how likely they are, and what ordinary people can do. In Farsi, with an English transcript.
Notes from an ICML 2026 anti-debate with Ludwig Schmidt on whether we should keep funding a human research workforce once AI is better at research than we are.
A year of notes on metabolism, hormones, food noise, and five months on a low-dose GLP-1 — for the women who've been told the only fix is more hormones.
Why privacy is contextual, why scaling won't fix it, and what we've learned from ConfAIde and CIMemories.
How the year I was on the job market became the year I built the healthiest habits I still keep today.
What types of postdocs are out there, and advice on how to decide if a postdoc is right for you.