Engineering — March 15, 2026
Reducing legacy platform load time from 4s to 600ms
Query batching, indexes, Redis, and a strangler-fig backend swap — same UI for dispatchers.
Read article →How we actually solved problems — not thought leadership.
These are write-ups from work we've actually shipped, not marketing filler. When we fix a slow query, swap a backend without breaking the UI, or set up a CI/CD pipeline, we document what we tried, what didn't work, and the numbers before and after. If a post doesn't have a specific technical detail in it, we don't publish it.
Query batching, indexes, Redis, and a strangler-fig backend swap — same UI for dispatchers.
Read article →Custom slugs, kill switches, expiry, and an API — what we built into our URL shortener and why.
Read article →Test gates, staged deploys, rollbacks, and audit trails — the minimum before you call it done.
Read article →Composite keys from slow-query logs, covering indexes, and when a fourth index makes writes worse.
Read article →If the screens already work, we usually swap the backend — not the UI. When a rewrite is actually justified.
Read article →NAT, snapshots, debug logs, and oversized databases — why the invoice jumps after you leave shared hosting.
Read article →Expand-then-migrate schema, feature flags, and a 15-minute abort path we actually rehearse in staging.
Read article →