Our game's Sentry had 324 unresolved issues. So many that real regressions were invisible.
One session with Claude and the Sentry MCP: 324 → 48.
How we sorted them 👇
3/ The other half: Seer autofix. Connected to GitHub, it reads the issue together with our logs and spans, drafts a fix and opens a PR.
So sometimes an issue arrives with the fix already waiting for review.
I show both in this video. Sentry sponsored it, and I use it every day in Tiny Harvest: youtu.be/6aBvhmaLqug
On the way to our Easter vacation, my dad messages me: something's wrong with the game.
Players were in the wrong co-op. Profile screens showed another player's data.
I hadn't touched the code in days. For an hour I was sure we'd been hacked 🧵
3/ The part that was on me: our API never told caches "don't store this".
Any response that depends on who's asking should send Cache-Control: no-store. Ours now sends it on every API response. One header, and an upstream cache mistake can't leak player data.
4/ Full story, plus three other things that almost killed our game: youtu.be/6aBvhmaLqug
My app pays my bills.
Over the last year it almost died 4 times:
1. Migrating app ownership
2. A CDN issue that became a nightmare
3. An email I did NOT expect
4. Losing the database the whole game runs on
None of them were bugs I wrote. 🧵👇
What happened, how I got out of it, and what I'd do differently if I started over:
youtu.be/6aBvhmaLqug
Thanks to @sentry for sponsoring. I use it in Tiny Harvest every day so I find out something broke before my players do. In the video I show how its AI helps me debug real errors from the game.
dub.sh/sentry-grimm


