The Daily
A note a day on APIs and the realities of shipping web applications. Stuff I've actually learned across 25 years of doing the work.
GitHub Actions artifact storage will quietly eat your budget
Nobody warns you that artifact storage in GitHub Actions accumulates forever by default. I learned this the hard way watching a client's bill climb for months.
Artifact storage in GitHub Actions is free until it isn't, and the transition happens gradually enough that you don't notice until you're staring at a GitHub invoice that's crept up three months in a row with no obvious cause. I hit this with a client running a mid-sized Laravel monorepo. Solid team, good test…
Read the daily →Covering Indexes vs. Composite Indexes in MariaDB: Not the Same Thing
I spent years conflating these two concepts. Once I understood the difference, I rewrote three schemas and shaved real milliseconds off production queries.
I spent an embarrassing number of years thinking "composite index" and "covering index" were two names for the same concept. They're not. Once I actually internalized the difference - and more importantly, once I started reading output with that distinction in mind - I rewrote parts of three different schemas and saw…
Read the daily →MariaDB Auto-Increment Locks Will Bite You at Scale
InnoDB's auto-increment locking modes are invisible at low volume and painful at high write throughput. Here's what I learned the hard way.
I spent a frustrating afternoon last year chasing insert latency spikes on a MariaDB instance that had no obvious slow queries, no missing indexes, and a perfectly reasonable schema. The culprit was , the default on some older MariaDB builds, combined with a bulk-insert pattern that nobody had thought twice about…
Read the daily →Claude's Token Counter: Why Your Estimate and Your Bill Disagree
Client-side token estimation for Claude drifts more than you'd expect. Here's why it happens and how to use Anthropic's actual counting endpoint to close the gap.
I assumed token counting was a solved problem. It isn't - at least not if you're trying to predict your Claude bill before a request hits the wire. I built a document processing pipeline for a biotech client last year. We were chunking lab reports, running them through Claude to extract structured data, and I had a…
Read the daily →R2 CORS Wildcards Don't Work Like S3 and It Will Break You
Cloudflare R2 CORS configuration looks identical to S3 on the surface. It is not, and the difference will silently kill your presigned upload flow.
I spent half a day last month debugging a presigned upload flow that worked perfectly against S3 for two years and broke the moment I pointed it at R2. The bucket was configured, credentials were valid, the presigned URLs generated fine. The browser just kept throwing CORS errors and I kept staring at a CORS config…
Read the daily →