// blog

Engineering notes_

Reliability, streaming, billing, and building with AI. The deep-dives start with case studies from real systems — more posts are in the pipeline.

featured deep-dive billing

Designing a webhook pipeline that survives duplicates, retries, and crashes

Billing webhooks are hostile input. How idempotency, deduplication, a canonical event model, and the Transactional Outbox pattern keep subscription state correct no matter what providers throw at you.

case study deep-dive · 8 min read → observability

Rolling out the LGTM stack: what actually shortened our debugging

Correlating logs, traces, and metrics by trace id — and why dashboards should be designed per audience.

case study deep-dive · 6 min read →
streaming

LL-HLS in production: latency budgets and the knobs that matter

Segment durations, part sizes, CDN behavior — where low-latency streaming is actually won or lost.

đang viết · coming soon
billing

Entitlements are a state machine: modeling subscriptions across Apple, Google, and Stripe

Grace periods, billing retry, upgrades mid-cycle — one canonical model to survive all three providers.

đang viết · coming soon
ai-assisted

My AI-assisted workflow: specs in, production code out

How I split work between human architecture and AI implementation — and where review effort really goes.

đang viết · coming soon
© 2026 Cao Hữu Phúc get in touch →