Matt Pocock
A prominent TypeScript educator who founded the Total TypeScript platform. He is known for his in-depth, practical teaching approach and is trusted by leading engineering teams worldwide.
Core Facts
Timeline (last 90 days)
Matt Pocock's old four-stage process includes the grooming skill, the two-step skill, the two-ticket skill, and the implement/code review skill.
The above content about the new skills mainly comes from a single video source, and the specific implementation details of Matt Pocock's original skills (such as prompt wording and parallel worktree configuration) are not elaborated in the video.
The retro skill reviews the agent's full conversation record and improves in two directions, context engineering and guardrails: generating agent MD files, cleaning up redundant system prompts, and codifying recurring errors into lint rules or coding conventions.
The PR skill changes the PR description structure to three sections: evidence, risk, and merge danger.
The key constraint of the two-ticket skill is to split by feature rather than by layer, so that each ticket can be completed independently by a single agent and the context window is protected.
A Bilibili creator believes Pstack's planning skill is better suited to large migrations and legacy refactoring, and that independent developers building from scratch should refer to Matt Pocock's or the Superpower series' planning skills.
Influenced by OpenAI researcher Sean Grove's view that "specs are the new code," Pocock tried spec-driven development for several months, but code quality kept deteriorating, and he eventually abandoned the approach.
Pocock believes that the essence of prompt engineering is not phrasing but precise conceptual modeling.
Pocock designed a skill called "grill me" that makes the Agent ask the user a series of follow-up questions in reverse, to dig out requirements and constraints the user is not aware of (unknown unknowns).
Pocock believes that the ability to communicate with Agents is becoming the core variable distinguishing excellent developers from ordinary ones.
40 more timeline events
All Facts (20)
Matt Pocock is the creator of the Total TypeScript course and previously served as a developer advocate at Vercel.
80%VerifiedMatt Pocock released v1.3 of his skills repository, introducing three new skills: implement-spec, PR, and retro.
65%UnverifiedMatt Pocock曾在Stately（XState背后的公司）工作，是XState的核心贡献者
90%UnverifiedMatt Pocock在成为工程师之前曾是一名声乐教练
85%UnverifiedThe retro skill reviews the agent's full conversation record and improves in two directions, context engineering and guardrails: generating agent MD files, cleaning up redundant system prompts, and codifying recurring errors into lint rules or coding conventions.
50%UnverifiedThe above content about the new skills mainly comes from a single video source, and the specific implementation details of Matt Pocock's original skills (such as prompt wording and parallel worktree configuration) are not elaborated in the video.
50%UnverifiedThe key constraint of the two-ticket skill is to split by feature rather than by layer, so that each ticket can be completed independently by a single agent and the context window is protected.
50%UnverifiedThe PR skill changes the PR description structure to three sections: evidence, risk, and merge danger.
50%UnverifiedMatt Pocock's old four-stage process includes the grooming skill, the two-step skill, the two-ticket skill, and the implement/code review skill.
50%UnverifiedA Bilibili creator believes Pstack's planning skill is better suited to large migrations and legacy refactoring, and that independent developers building from scratch should refer to Matt Pocock's or the Superpower series' planning skills.
50%UnverifiedPocock designed a skill called "grill me" that makes the Agent ask the user a series of follow-up questions in reverse, to dig out requirements and constraints the user is not aware of (unknown unknowns).
50%UnverifiedPocock believes that software engineering fundamentals have become more important in the AI era, because an Agent's operating environment is the codebase itself, and a garbage codebase leads to garbage output.
50%UnverifiedPocock believes that the essence of prompt engineering is not phrasing but precise conceptual modeling.
50%UnverifiedPocock believes that the ability to communicate with Agents is becoming the core variable distinguishing excellent developers from ordinary ones.
50%UnverifiedPocock suggests that the direction for people without a technical background to seize the AI opportunity is to bring Agents into knowledge work processes and bring those processes to their organizations.
50%UnverifiedPocock suggests practitioners follow a "trust ladder": first use Agents on the tedious work you hate most and keep observing, then gradually expand starting from simple tasks that are measurable and correctable.
50%UnverifiedPocock believes that the effectiveness of a methodology depends on the quality of the underlying environment: however detailed the spec, if the codebase itself is chaotic, the Agent will still produce more chaotic results.
50%UnverifiedPocock believes that the reason code is the first domain Agents have conquered is that code is easy to verify; real-world knowledge work has much higher verification costs, which is the main bottleneck for current Agent deployment.
50%UnverifiedPocock currently runs four Agents simultaneously handling long-term tasks for different projects, with about 200 pushes per day on GitHub.
50%UnverifiedMatt Pocock believes the planning phase requires deep human involvement while the coding phase can be safely delegated to AI; starting to write when requirements are vague leads to rework costs far higher than upfront alignment.
50%