Add new blog post about the history of AI #watchdog

Make a sale from a random guy after the last customer churned - this brings me back to $126 MRR (up from $97) 🎉 #watchdog

Add metadata to each persisted chat completion so I can filter and group training data by community and other props #watchdog

Enable OpenAI's new "store" param to automatically persist chat completions for 30d so I can create fine-tuning jobs more easily in the future #watchdog

Ship automated testing feature and tweet + linkedin post about it: x.com/ben_makes_stuff/status/… and www.linkedin.com/feed/update/… #watchdog

See test case runner working during my integration testing with real data, very cool #watchdog

Finish testing "create" and "get" flows for test case management - working! Now just need to finish testing the remaining operations + "import from rule violation log", pushing to finish by the end of the day today #watchdog

Finished CRUD APIs for test case management, now to test everything with the frontend and make sure it works #watchdog

Add a bunch of SQL queries to handle test case creations, updates, etc #watchdog

Write, execute, and verify some fairly complicated SQL to backfill rule ids that were violated now that I need to use those in addition to the display names for the backtesting feature #watchdog

Write some fairly complicated SQL to backfill rule ids that were violated now that I need to log those in addition to the display names for the backtesting feature #watchdog

Add database migrations for the backtesting feature for Watchdog #watchdog

See first situation happen where a customer churned and ask them what I could have done better, back below $100 MRR 😬 #watchdog

Finished the "edit test case" UI and got pre-populated fields working properly #watchdog

Finish "create test case" UI work and integrate with rules endpoint to fetch the list of allowed options #watchdog

Start handling OpenAI "refusals" because sometimes gpt-4o ends up refusing to analyze images for whatever reason and I'm getting paged. This is a short term solution to stop getting paged, medium term solution: will add fallbacks to different providers #watchdog

Build "log -> test case" import UI www.youtube.com/watch?v=OaNfV… - cc @marc added a simple beaker icon you can tap to import a test case from a given log #watchdog

Moved rule violation tags inside each log card instead of awkwardly sitting on top outside the card (old behavior). Think I like this UI better #watchdog

Start building the saved test runner for Watchdog so customers can more easily experiment with rule wordings without worrying about regressions: www.youtube.com/watch?v=flFaH… - h/t @marc for the idea #watchdog

Fix rule table alignment by using table-cell instead of block for th and td #watchdog