2026-06-02
Latency budgets: designing ML inference you can defend
P99 is a contract. How we set latency budgets before training, and the serving architecture that keeps them.
Read dispatch ↗
[ // ] — DISPATCHES
Latest dispatches
Field notes, migration playbooks and engineering evidence from the studio.
2026-08-14
Zero-downtime Odoo migrations: the rehearsal method
How we move 40M+ records between Odoo majors without a maintenance window: dual-run, reconciliation gates, and a reversible cut-over.
Read dispatch ↗