- ai
- building
- devtools
Google Made Flash 4x Faster and Changed Our Latency Budget
A feature we shelved in September because 900ms TTFT felt dead came back off the shelf eight months later when Gemini 3.5 Flash hit 210ms in production.
- building
- ai
- infra
We Rebuilt Our RAG Pipeline Three Times. The Third One Used Postgres.
Two failed approaches, one production outage during a customer call, and the realization that the synchronization problem disappears when you stop adding databases.
- ai
- devtools
- postmortem
We Ran Four AI Agents at Once for a Week
Cursor 3 shipped the Agents Window and we immediately learned the hard way what parallel agents actually need: exclusion lists, worktrees, and honest accounting of what gets traded away.
- ai
- typescript
- devtools
Pin Your AI Models Like You Pin Your Dependencies
Vendors silently roll model versions and your production output drifts. Lock the version, log the cost, ship a fallback. The playbook in TypeScript.
- ai
- typescript
- security
Giving an AI Agent Permission to Spawn Sub-Agents (Without Losing Control)
Scoped delegation with cascading revocation, depth limits, and per-agent audit trails in TypeScript. Concrete code, not just theory.
- ai
- postmortem
- devtools
Claude Code Felt Off for a Month. Here Is What Broke.
Three overlapping bugs caused four weeks of subtle Claude Code regressions that no standard test suite would have caught.
- security
- devtools
- ai
An AI Tool Had OAuth to Their Whole Google Workspace. Then Vercel Got Breached.
A Vercel employee clicked Allow All on an AI app's OAuth consent screen. Three weeks later, customer environment variables were on a hacker's drive with a two-million-dollar asking price.