@tobias
One integration puts hundreds of tools in reach. Here is everything we had to build around it before it was safe to point at a real inbox.
August 31, 2026 · 7 min read
Two months of latency work on a production chat agent: the wins, the ideas the measurements killed, and the numbers behind both.
August 24, 2026 · 6 min read
A 100% cache hit makes tokens cheaper, not invisible. Deferring the tool contracts halved our chat's time to first token.
August 12, 2026 · 2 min read
Every harness bug masquerades as a model result: false reds read as regressions, false greens read as coverage.
July 26, 2026 · 2 min read
Fanning out parallel reviewers catches confident wrong answers. It also produces confident wrong findings.
July 22, 2026 · 2 min read
Replacing a plaintext MCP secret with OAuth, publishing the SDK, and moving a live site off Sanity.
July 21, 2026 · 6 min read
July 21, 2026 · 2 min read
July 21, 2026 · 2 min read
July 21, 2026 · 2 min read
July 21, 2026 · 2 min read
July 21, 2026 · 2 min read