Curated internet, served daily
Models, agents, and what actually runs
Models, agents, and the tools built on top of them — what runs, on whose hardware, at what cost. I care about the gap between a benchmark and a Tuesday, so most of what lands here is something I ran rather than something I read about.
Two old bug classes — a heap overflow and an SSO misconfiguration — carried a paid security team into OpenAI's internal repositories in under 72 hours, and the identity mistake is the part that decided how far the breach could travel.
Cloudflare's Clairic claims faster, cheaper, more accurate extraction of unstructured data than general LLMs. The claims are unverified, but the narrow-model bet is worth a test.
A forkable Claude Code framework that evaluates postings, tailors CVs and preps interviews on a local machine — the editable pipeline is worth more than the letters it writes.
Gemini 4 Argon's launch is mostly benchmarks and partner news; the one capability worth watching turns raw video into simulation code that estimates depth and boundaries.
Anthropic researchers say small models are months from self-replicating on smart ovens and off-brand TV boxes — the risk sits on cheap, unpatched hardware nobody updates.
A comparison of Claude Opus 5.5 and GPT-6.1 Sol subscriptions finds Anthropic's plans about 5x ahead on API-equivalent value for agent work.