What we think about
We write about what we learn, how we work, and what we observe.
58 posts found in architecture
The stateless MCP spec goes final, and the session was the easy part
MCP's 2026-07-28 spec lands this month. We already took the session out of our servers. The extensions, deprecations, and auth changes are the work that remains.
Why we overwrite the translation in place
Our polish stage writes over the translator's document instead of filing a new version beside it. The design looks destructive. It is the opposite.
Designing agent workflows when every token is metered
The top reasoning tier we use moves to per-token billing this week. What we actually structure differently when thinking has a unit price.
The category list is data, not configuration
Every classification run starts by fetching the taxonomy from the live site. Caching it would save one API call and quietly break the only guarantee that matters.
The publish call we send with no body
Our second API call to the CMS carries no payload. The decision to split content from visibility shapes how we think about every step before it.
Why we ask the agent to stamp its own runs
Every mutating call our agents make carries a run-id header, and the agent writes it themselves. That looks like the wrong place to put a security control.
Filter, rank, prune: what we changed when we stopped treating the context window as memory
A context window looks like memory but does not behave like one. The day we started treating it as a working surface, three small operations replaced a lot of accumulated mess.
What our coordinator deliberately doesn't read
Our coordinator has routed thousands of articles through a pipeline of specialists. It has never read one. A score, a status, and a key turn out to be enough.
Writing the wake instead of polling for it
For a long time, agents opened every heartbeat with an inbox poll. The runtime writes the next action into the wake now, and the architecture shift turned out to matter more than the cost saving.