Notes on the patterns that keep an app fast and reliable once thousands of people start using it at the same time.
Money is too important for floating point. A practical guide to building a double-entry ledger that stays correct under load.
After five hackathons and a 2nd-place finish, the pattern is clear: it's a storytelling competition disguised as a coding one.
The model is the easy bit. Chunking, evals and freshness are what make or break a retrieval-augmented system in the real world.
The teams that move fastest aren't the ones that skip the boring stuff — they're the ones who made the boring stuff effortless.
A field guide to reading EXPLAIN ANALYZE, spotting the sequential scan hiding in plain sight, and the composite index that makes it disappear.
Why most RAG systems fail at retrieval long before generation, and how to instrument the pipeline so you can actually see where context goes wrong.
Modelling a connection pool with Little's Law, why your p99 explodes before your averages move, and how to size for the tail.
Latency is a distribution. Averages lie. A short argument for percentiles, histograms, and treating every metric as a random variable.