Matt Pocock and Nate Herk explore the communication gap between a person and an agent. Better instructionsContext engineering designs the information, instructions, memory, and tool state an AI receives so it can perform a task reliably. begin with understanding what matters, asking questions that expose missing requirements and agreeing on the terms used by clients, users and code.
Matt Pocock describes learning from agents while checking high-quality primary sources rather than treating generated explanations as authoritative. He connects precise language and domain-driven designDomain-driven design organizes software around the concepts, language and rules of the business or problem domain it serves. to the practical work of making a system easier to change and verify.
Matt Pocock recounts a specification-driven experiment that did not work well for his own video editor. Refocusing on code quality, tests and an environment the agent could actually run improved his results. This is his experience, not a controlled comparison proving that specification-driven development always fails.
Matt Pocock and Nate Herk recommend building trust gradually through observable, measurable tasksAgent completion verification checks observable evidence that an agent achieved the requested result instead of accepting its claim that the work is finished. and human review. Their claims about personal productivity, model preferences and future software factories remain experience and speculation rather than universal performance guarantees.
Watch on YouTube
