Potato
A former Meta React team engineer, currently at Cursor (SpaceX AI), focused on large-scale AI agent engineering practices. Known for two methodologies: the "Skill Library" approach, which systematizes reusable agent capabilities, and the "Trust Ladder" framework, which governs the progressive delegation of tasks to AI agents.
Timeline (last 90 days)
Stack's author goes by the online name Potato (real name Laurentan), has Meta and Netflix on his résumé, later joined Cursor as a Member of Technical Staff, and is also a React core team member involved in maintaining React Compiler.
The entire Stack plugin has no skill that does planning; the author does not believe in planning before acting, and holds that the best specification is the code itself.
The critical first step to climbing the trust ladder is giving agents verification capabilities ("hands and eyes").
Potato believes that as models become increasingly powerful, the bottleneck is no longer the agent itself but rather humans' ability to clearly articulate intent.
A former Meta React team engineer named Potato, now working at Cursor, claims to have merged 2,500 PRs to production last month.
Potato built an internal framework called Dune for the team's Electron application, featuring a large number of strict lint rules so that there is only one correct way to do any given thing.
A large portion of the 2,500 PRs consisted of ongoing maintenance work that Potato calls "gardening," and were not 2,500 new features.
Potato mentioned that he likes Opus 5.5.
The engineer's remarks about 2,500 PRs garnered approximately 3 million views on X.
All Facts (5)
Stack's author goes by the online name Potato (real name Laurentan), has Meta and Netflix on his résumé, later joined Cursor as a Member of Technical Staff, and is also a React core team member involved in maintaining React Compiler.
50%UnverifiedThe entire Stack plugin has no skill that does planning; the author does not believe in planning before acting, and holds that the best specification is the code itself.
50%UnverifiedThe critical first step to climbing the trust ladder is giving agents verification capabilities ("hands and eyes").
50%UnverifiedPotato believes that as models become increasingly powerful, the bottleneck is no longer the agent itself but rather humans' ability to clearly articulate intent.
50%UnverifiedPotato built an internal framework called Dune for the team's Electron application, featuring a large number of strict lint rules so that there is only one correct way to do any given thing.
50%