ThesisMay 6, 2026·6 min
The trust grammar: how we think about AI accountability.
Approvals aren't friction. They're the contract that makes AI as a teammate possible. Notes on building trust into the architecture, not bolting it on after.
Coming soonThesisApr 30, 2026·5 min
Intelligence isn't artificial. It's social.
The thinking we actually want from AI doesn't happen inside one mind. It happens between minds — with the right context, the right people, in the same room.
Coming soonThesisApr 22, 2026·7 min
What Privacy First AI actually means.
Without invisible memory, hidden actions, or uncontrolled automation. Walking through the operational commitments that turn a marketing phrase into an enforceable contract.
Coming soonBuildingApr 14, 2026·11 min
The four planes of orchestration.
Temporal for durable workflow. LangGraph for in-run reasoning. Celery for async execution. RabbitMQ as transport. The engineering choice that took the longest to defend — and is now load-bearing.
Coming soonBuildingApr 5, 2026·9 min
Shipping an approval card.
A close look at one component: the approval card. Why every shipped UI element should be defensible in a room. What we changed twice. What we won't change again.
Coming soonBehind the scenesMar 28, 2026·4 min
Why we call it Early Access, not public beta.
Beta is a word that's stopped meaning anything. Early Access, the way we use it, means something specific: invite-only, rolling, personal onboarding, real product — not "we wrote a TODO list and called it ready."
Coming soon