Lauren
An engineer on the Cursor team focused on AI agent collaborative engineering practices. Lauren proposed the trust curve model and verification skill framework, and developed the open-source PSAC plugin.
Timeline (last 90 days)
Lauren said: "If you only have rules and BugBot, the codebase will sooner or later turn into garbage."
Lauren created a "Feature Map" document that teaches the Agent how to reach all features, including user-perspective paths, keyboard shortcuts, and the DOM elements and attributes used for CDP selection.
Lauren built a skill called "Control Glass" that lets the Agent control Electron apps via the Chrome DevTools Protocol and capture traces.
Lauren believes Verification is the most important skill for collaborating with Agents.
Lauren worked at Cursor for five months, and in the first month his output was extremely low due to unfamiliarity with the codebase.
Former Cursor engineer Lauren shared his AI Agent collaboration practices on Bilibili: from initially watching a single Agent closely, to now letting Agents automatically merge PRs.
Lauren spent over 600 PRs refactoring GrokBot into the "Dune" architecture, and now no longer looks at the code.
Lauren believes Brownfield applications are safer than Greenfield, because big tech companies design their infrastructure for "the least skilled engineer," and Agents benefit from this equally.
Lauren believes every comment left under a PR saying "this shouldn't be done here" should be treated as a code smell, and the right approach is to turn the constraint into a hard rule, a lint rule, or a CI failure.
Lauren built two skills, 'how skill' and 'why skill,' enabling the harness to more rigorously examine the codebase and proactively investigate the codebase for follow-up details upon receiving a Slack message.
30 more timeline events
All Facts (20)
A Feature Map is a special file bundled with a skill that teaches the Agent how to navigate to every feature of the product from the user's perspective, including sub-features, keyboard shortcuts, and the data attributes of CDP-selected elements.
60%UnverifiedLauren said: "If you only have rules and BugBot, the codebase will sooner or later turn into garbage."
50%UnverifiedLauren created a "Feature Map" document that teaches the Agent how to reach all features, including user-perspective paths, keyboard shortcuts, and the DOM elements and attributes used for CDP selection.
50%UnverifiedLauren built a skill called "Control Glass" that lets the Agent control Electron apps via the Chrome DevTools Protocol and capture traces.
50%UnverifiedLauren believes Brownfield applications are safer than Greenfield, because big tech companies design their infrastructure for "the least skilled engineer," and Agents benefit from this equally.
50%UnverifiedLauren believes Verification is the most important skill for collaborating with Agents.
50%UnverifiedLauren worked at Cursor for five months, and in the first month his output was extremely low due to unfamiliarity with the codebase.
50%UnverifiedLauren spent over 600 PRs refactoring GrokBot into the "Dune" architecture, and now no longer looks at the code.
50%UnverifiedLauren believes every comment left under a PR saying "this shouldn't be done here" should be treated as a code smell, and the right approach is to turn the constraint into a hard rule, a lint rule, or a CI failure.
50%UnverifiedLauren built two skills, 'how skill' and 'why skill,' enabling the harness to more rigorously examine the codebase and proactively investigate the codebase for follow-up details upon receiving a Slack message.
50%UnverifiedLauren developed a plugin called PSTACK, which is a set of skills encoded as markdown documents, inspired by GaryStack from YC CEO Gary Tan.
50%UnverifiedLauren joined Cursor six months ago to work on optimizing performance issues in the Cursor Agent window.
50%UnverifiedLauren built a plugin named P-Stack, which is essentially a collection of skills that distills her years of software engineering workflows into corresponding playbooks and skills.
50%UnverifiedLauren advises against relying solely on style guides, noting that manual review is only suitable for catching what is missing, and recommends investing time in the earlier layers instead.
50%UnverifiedLauren believes the biggest obstacle to transitioning from a small number of agents to a large number of agents is not compute or tooling, but trust.
50%UnverifiedThe trust ladder proposed by Lauren, ranked from highest to lowest effectiveness, is: the codebase itself, static analysis, rules/Bug Bot/skills, and style guides.
50%UnverifiedWhen an Agent repeatedly makes the same mistake, rather than correcting it each time, it is better to add a lint rule or refactor the codebase so that the mistake is structurally impossible.
50%UnverifiedLauren previously worked on the React team and joined Cursor (later merged into SpaceX AI) about half a year ago.
50%UnverifiedLauren believes that the core of an efficient Agent workflow is built around 'trust,' and engineers need to instill knowledge into the Agent to avoid becoming a bottleneck themselves.
50%UnverifiedForcibly launching hundreds of sub-Agents in parallel in the absence of trust will generate a massive volume of junk PRs, regressions, and bugs.
50%