Every Test Passed, and There Was Still a Bug: Finding Ordering Bugs with pstack and TLA+We installed pstack, the agent skill set a Cursor engineer uses every day, into Claude Code and used formal verification (TLA+) to find and fix a bug the tests missed. Then we opened the open-source code line by line and matched "this part works like this, so the agent did that" against the real session logs.
Dante (Quokka Labs)Oct 24, 2026 
