Building thoughtful software with AI.Notes from the systems behind the work.
An incremental sync that reset its coverage on every shape change, and starved its own connection pool, is rebuilt around batched transactions, a serial write queue, and a separate refresh cursor.
Read moreThe button that saves a delivery as a draft in the stores' inventory system checked first and wrote its record only after a slow outside call, so a second click could slip through. Now it claims the delivery in the database first, lets a partial unique index refuse the second claim, fixes a payload hash that couldn't see nested fields, and adds a narrow admin cancel for claims that get stuck.
The Unspecified card on my discount report filtered for the literal word instead of the rows with no supplier, which SQL can never match. A shared marker and one filter builder fixed the drill-down, with a few gaps I can still see in the diff.
The save at the end of my ordering wizard catches each failed item, logs a warning and keeps going inside one transaction. Postgres aborts the transaction at the first database error, so the partial save it promises can't happen, and the same file reloads sessions by timestamp and drops leftover units.
The agent built an encrypted, signed, two-copy backup system with no tests. Reading the restore path showed a loose success flag, a signature nobody checks, truncate-before-load, and comma-split CSV, plus what I want restore to do instead.
An admin endpoint runs signed, encrypted SQL uploads, and a dry run can reset a version it already applied.
Turning chores into templates forced a real upgrade path that asks the database what it has before adding anything. Reading the diff, I found a second route to the wipe still open and a backfill that can get stranded.
A post-import step meant to backfill analytics views drops and replays each view's stored definition, which restores the trigger but not the rows, touches every org's views, and swallows failures, all to fix a gap nobody ever counted.
A few of my ClickHouse rollups kept an average next to the total, and the summing engine would have added those averages on every merge; now they keep total and count and divide at read time, and the affected rollup keys now include every label they store.
Some of my databases name the lookup key column differently, and since SQL won't bind a column name, I pick it from a two-item list in the code by asking the schema, then let each caller decide what a miss means.
A small retrieval service guarded its saved data with create-if-not-exists while its text table changed shape twice in one day, so an older store would likely fail on missing columns, and the obvious drop-and-recreate could point saved vectors at the wrong rows.