Lauren Tan

Software engineer and the creator of pstack, the agent skill stack built from the workflows she uses to ship code. Her X handle is @poteto. The auto-generated captions of 2026-10-05-youtube-poteto-software-factory garble the handle as “potato” and once spell it “potato-noodle”; the oEmbed title and the stream description both say poteto.

Career

Her LinkedIn bio reads “Building Grok Bot and Cursor at SpaceXAI. Previously I worked at Meta and Netflix,” and separately that she is on the React Core team. The stream corroborates the shape and fills in the timeline: she worked on the React team at Meta, took a month off when burnt out and started the side project that seeded pstack, joined Cursor in March, worked on the laggy agents window (5:20 to 5:41), and later refactored the Grok Bot codebase into an Electron architecture she calls Dune (30:10). She managed a team of UI engineers at Netflix, and cites the manager idea “context not control” (41:40). Her GitHub profile https://github.com/poteto reads “Software Engineer @xai-org” alongside the React compiler core team; the employer chain beyond the captions is firm from these profiles, not caption-derived.

What she argues

Her central claim is an inversion (7:17): as models get better, “the bottleneck is no longer the agent,” it is your ability to express intent so the agent can carry it out. From that follow the core pages this stream files: trust-ladder-agents, agent-verification-skill, determinism-extraction, environment-and-constraints, and inner-outer-loop-agents. She says she ships roughly 2,000 to 2,500 pull requests to production a month with high confidence, the higher figure the one she cites on this stream and the lower the figure in her own Pt. 1 guide, and that a lot of them are gardening rather than features (46:01). On air she studied drama and returns often to language as the tool, agreeing with the host that a skill is just process turned into words, skills-as-process.

Sources