Posts

All the articles I've posted.

• Your LLM can't hallucinate a number it was never given

  18 Jul, 2026·3 min read

  Preventing invented facts in the code that builds the prompt rather than in its instructions: no raw numbers, no empty fields, and opaque references instead of names.

• Cache the prompt, not the inputs

  27 Jun, 2026·2 min read

  An LLM cache keyed on a hand-picked list of inputs fell behind the prompt it protected. Hashing the rendered prompt fixed it, and the choice of hash turned out to be a security decision.

• Fire-and-forget doesn't survive on Cloudflare Workers

  5 Jun, 2026·3 min read

  A view counter returned 204 on every call and never moved. Workers can cancel work still pending after the response unless it is passed to waitUntil.

• What a non-extractable key actually protects

  30 May, 2026·4 min read

  Keeping a signing key in the browser under a WebCrypto key that cannot be exported: what that stops, what it does not, and where the real protection comes from.