Building thoughtful software with AI.Notes from the systems behind the work.
When a spreadsheet row didn't match a supplier's own catalog, my import searched every product and wrote case sizes to whatever it found, and the schema let two suppliers share a product. I deleted the wide lookups, put an ownership check and a preflight in front of the writes, moved one-supplier-per-product into the database, and left a few loose ends I'm still not happy about.
Read moreWiring two issue trackers into my dev-metrics app got the handshake signed and the tokens into a vault, but disconnect now has to delete webhooks before the token that removes them, and the inbound event handlers still trust whoever calls.
The lookup that decided whether an organization already had a payment customer searched by an email my code sometimes guessed or generated, and could hand one org's billing record to another. I moved identity onto the identity provider's org ID and tied retries to it, but the same commit left an ignored retry flag, a delete check that reads error wording, a gutted seat counter, and no tests.
The per-tool cost chart now reads two typed views instead of parsing every raw log row, the migration that builds those views errors on a leftover instead of skipping it, the documented session join filters by organization on both sides, and the drift check covers more token types but still only warns.
An agent-facing tool in my observability service binds the tenant properly but hand-quotes an optional list of session IDs into the query text, so one model-chosen ID can turn the tenant check into one side of an OR.
A dashboard-as-code prototype drew its charts with the app's old chart code and printed the dashboard spec underneath, so editing the spec changed nothing; the real result was a version-conflict warning sitting under a row of green checkmarks.
The agent console's transcript store and its telemetry store got merged by timestamp. The duplicate check asks whether anything is nearby, not whether the matching thing is, so real tool results can vanish in bursts, and the join loosened the telemetry access filter along the way.
The admin staff list in my scheduling app sent pages with no total, so Next never switched off. The server now counts after the search filter and before slicing, caps the page size, and the screen does the page math; saves also refuse rows with no company, since those could never appear on any page.
The schedule parser built its name lookup from a staff table shared by every company, admins included. It now has to be told the company, both import endpoints check it against the token, and anyone outside that company comes back unmatched.
The scheduling app's backup page grew a required company picker, per-company folders and a path check in one day, but the step that makes the dump never got the company, so every file holds every customer. Restore has the same problem in reverse.
Opening the roles endpoint to company admins let them list, edit and delete the one role that wasn't theirs. Each operation names a role differently, so closing it took four checks, and the rename-by-id one is the one I'd have missed.