Blog

Prisma will happily turn one findMany into 133 statements, and the include that replaces them is not one query either, so this walks through the real logged SQL for four versions of the same read against a 132-post table.

The XOR and popcount tricks behind Single Number, Missing Number and Counting Bits are small enough to verify exhaustively, so this runs them over every input they claim to cover and reports the exact value at which each one stops being correct.

Runes look like function calls but compile into signals, proxies and effects, so the fastest way to understand them is to read what svelte/compiler emits for each one.

The moment a credential reaches a remote, every copy of it is already out of your reach, which is why rotation is the fix and deleting the line is theatre.

A single column rename carried through expand, dual-write, backfill, cutover and contract on a live Postgres 17 instance, with the lock mode and timing measured at every step.

The stack loop that evaluates an RPN expression is fifteen lines, but it returns undefined for an empty array, NaN for a missing operand and -0 for a negative fraction, so the real work is deciding what your evaluator owes a caller who hands it nonsense.

Four rate limiting algorithms, one request stream, and a harness that counts what each of them actually admits in the worst rolling second.

A list CRDT has one job — converge — so you can brute-force it: 400 operation logs, every one of the 720 delivery orders each, diff the results. Three orderings broke the naive implementation in three different ways, and one of them left every replica in perfect agreement about a scrambled document.

A Bloom filter's error rate is one of the few things we teach that you can actually check, so I built one, inserted 500,000 keys, queried it with ten million keys that were not in it, and compared the result against the textbook formula at seven different sizings.