Pulling sanitized prod data into dev when rsync isn't there
A rake task copies the production SQLite file down, nulls out every secret before it touches dev, and survives schema drift and a host without rsync.
A rake task copies the production SQLite file down, nulls out every secret before it touches dev, and survives schema drift and a host without rsync.
Mirroring the Pexels client line for line, a fire-and-forget download-tracking ping, and a browser-controlled URL that almost got the API key attached to it.
A routine provider-registry refactor turned up an unvalidated param that could commit a mislicensed image to the live blog as legitimately self-hostable.
A download-and-rehost pipeline for CC images: vips from a buffer, a UTF-8 force-encode that would corrupt JPEG bytes, and a stale queue mimicking a bug.
Closing the hero-image loop: insert-only frontmatter patching, a guard that caught real drift on its first run, and a Turbo broadcast with no listener.
One PUT endpoint, an error-mapping decision (409 is the optimistic lock, 422 is your fault), a PRD that pointed at a file that doesn't exist, and a salvageable stash on a dead branch.
Nine dead Postiz columns, a memoization where Rails.cache would have defeated the encryption, three Ruby filters moved into SQL, and unreachable code that wasn't.
Adding i18n pipeline plumbing to a Rails blog syndication app, and the partial-migration bug my own code review caught before it shipped.
Eight jobs, one broadcast skeleton, two twin poll jobs that diverged on the only line that matters - and a rescue-ordering bug caught by tracing, not tests.
A boolean threaded through four layers, and three independent reviews agreeing on a finding I deliberately didn't fix, since the cache is the dedup mechanism.
Page 1 of 6