Featured
Cutting API latency 9× with embedded DuckDB in Rust
How a read-only analytics API went from tens of seconds to low single-digit seconds at p99, and how the storage behind it shrank by roughly three quarters.
ENGINEERING NOTES
Lessons from production.
Short, practical write-ups on performance, data architecture, API design, cloud infrastructure, and AI in real products. Each one reads in about a minute.
11 articles
Lakehouse or OLAP cluster? Choosing storage for an embedded engine
Why DuckLake fits an embedded architecture and a separate OLAP cluster doesn’t, and the hidden risk of a catalog database on the hot path.
Adding search without replacing what works
Comparing built-in full-text search, an embedded Tantivy index, and a new database as a sidecar, with a staged rollout for each.
Fixing N+1 in a Node GraphQL service without breaking clients
Moving to Bun, adding DataLoaders, and keeping the output byte-for-byte identical.
Small AI models as classifiers, not chatbots
Using a schema-constrained model to pick the right query in an MCP tool, with permission-filtered options, keyword fallback and shadow-mode rollout.
A support assistant that explains instead of acting
Grounding a first version in a glossary and error pages so non-technical users get clear answers without the risk of an agent taking actions.
Access control belongs on the server
Why letting clients send their own filters breaks API security (OWASP API1/BOLA), and how to enforce entitlements in GraphQL.
Upgrading MySQL 8.0 → 8.4 on RDS: the blockers nobody mentions
Parser overflows in stored routines, the end of mysql_native_password, and privilege changes, with a checklist to test before the upgrade.
Honest service-health reporting
Separating real failures from rejected calls, setting error budgets, and treating “found by the weekly report, not an alarm” as a finding in itself.
AI agent teams across the software lifecycle
Ten role-scoped Claude Code agents, from requirements to release, with write access limited per role.
When timestamps lie: lessons from a large product catalog
A bulk reindex made updated_since useless for release-date queries, and high IDs turned out not to mean recent releases.