cs-core

Last quarter I helped a team debug a build pipeline that looked fine on paper: two runners, dozens of short jobs, and a clear rule that at most two jobs may be active at any moment. The surprise was a hidden assumption about boundaries. One job ended at 10:00, another started at 10:00, and the […]

Last month I was profiling a service that needed every prime up to a limit to build a deterministic hash ring. Trial division was fine for a single value, but it fell apart when I needed all primes <= 50 million. That moment reminded me why prime sieves still matter in 2026. When you already

Most of the systems I build that rely on ordered data eventually run into a small but annoying problem: I want sorted output without paying the cost of materializing the whole tree. If I store results in a BST for fast inserts, recursion-based traversal works, but it is awkward when I need to pause, resume,

A few years ago I inherited a payments SDK where every formatting variation had a different name: formatMoneyInt, formatMoneyDecimal, formatMoneyWithCurrency, formatMoneyWithLocale. The code worked, but every caller had to memorize the naming grid, and new variants kept piling on. The core idea was a single action—formatting money—yet the API spoke in four dialects. That kind

At 2 a.m. I was chasing a production outage in a payment pipeline. The logs were 200 MB, and I needed every line that contained a specific error signature. I could have opened the file and done a linear scan, but I wanted the result now, not minutes later. That moment is why pattern searching

Last quarter I had to schedule a pile of nightly data jobs on a small analytics box. Each job took a fixed one‑minute slot, and if it finished after its deadline the customer report was late and the revenue bonus disappeared. That real constraint maps perfectly to the classic job sequencing problem: pick a subset

Last year I helped a team triage a production outage in a payments platform. A single dependency loop between the fraud service and the ledger service caused retries to bounce forever. On the surface it looked like a timeout issue, but the real culprit was a cycle in the service call graph. That moment reminded

Last quarter I was building a permissions feature for a product taxonomy. Each category node carried an integer label that came from a legacy sync, and for every category I needed the smallest missing non‑negative label inside its subtree. That number became the next safe ID for automated inserts. Doing this once is easy; doing

Last month I was called into a checkout outage. A discount rule that sounded simple—apply a bulk discount only for physical items, except for gift cards, unless the customer is in a loyalty tier—was coded as a flat chain of if statements. The rule looked correct at a glance, but one edge case slipped through

Last winter I was on call for a payments API that had to check whether a transaction ID had been seen before. We stored the IDs in a sorted array because we also needed range queries for fraud analysis. The array was large enough that a linear scan was off the table, so I reached

Last week I was reviewing a scheduling system for a robotics lab when an intern asked how to build a magic square for a UI demo. That kind of request shows up in interview prep and algorithm classes, but also in procedural generation and test data. When n is odd, we have a deterministic method

Last quarter I worked on a ranking service that was correct but felt wrong. Scores were right, yet the order of items with the same score changed between requests. Users thought the list was unstable and filed bugs. The fix wasn’t a new scoring model; it was a sorting choice. We swapped from a fast