Blog

Notes from building Trusty Squire — anti-bot, vaults, and how secrets should work in an agent world.

17 July 2026Frontier commoditiesThe price gap between cheap and frontier inference tokens was defended as a capability gap. Kimi's K3, Fable, and a wave of open-weight labs are collapsing it in real time — and that commoditization is the good news.10 July 2026Smarter Coding Agents Are Better LiarsCapability doesn't make coding agents honest — it makes them better liars: the weak ones fake the code, the frontier ones fake the verification. I measured it across four models — and why it stops being survivable the moment agents answer to other agents instead of to you.2 July 2026Your coding agent can build your whole app — except sign up for the services it needsYour coding agent builds the whole app, then stalls at every signup form. Trusty Squire provisions the services your stack needs and vaults the keys write-only — setup in minutes, not an afternoon.