Writing · Nº 041065 essays
The writing
Ideas from inside the work. Explore technology, entrepreneurship, and the practice of building things that last.
15 essays found · Page 1 of 2
Clear filtersEngineeringMake AI Agents Ask Questions Before They Start CodingGetting AI agents to ask questions before coding kills a whole class of wasted work. Here is how to make an agent clarify ambiguity instead of guessing wrong.EngineeringTDD With AI Coding Agents: Tests as the SpecTDD with AI coding agents works better than with humans because agents excel at making a red test green. Here is how to use tests as the spec the agent codes to.EngineeringWhy Long AI Coding Sessions Degrade, and How to Fix ItLong AI coding sessions degrade as the context fills with noise. Here is why an agent gets worse the longer it runs, and how to reset it before it does.EngineeringAI Code Review vs Human Review: Where Each WinsAI code review vs human review is not either-or. Here is what an AI reviewer catches, what only a human catches, and how to run both without wasting either.EngineeringHow to Teach an AI Agent Your Codebase ConventionsTeaching an AI agent your codebase conventions is how you stop fighting its defaults. Here is how to encode your patterns so agent output matches your repo.EngineeringHow to Scope an AI Coding Task So It Actually SucceedsScoping an AI coding task right is the difference between a clean diff and a mess. Here is how to size and bound a task so the agent finishes it correctly.EngineeringWrite Characterization Tests Before You Let AI RefactorCharacterization tests before an AI refactor pin down current behavior so you can tell a clean change from a regression. Here is how and why to write them.EngineeringWhen to Throw Away AI Code and Start OverKnowing when to throw away AI code and restart saves more time than debugging it. Here are the signals that mean stop fixing the agent's output and reset.EngineeringA Definition of Done for AI-Assisted TasksA definition of done for AI-assisted tasks keeps agents from stopping early or gold-plating. Here is the checklist I use so done means done, not looks done.EngineeringShould AI Write Its Own Tests?Should AI write its own tests? Sometimes, but not the ones that verify it. Here is the split I use so agent-written tests help instead of hiding bugs.EngineeringA Test Strategy for AI-Generated CodeA test strategy for AI-generated code has to catch different failures than one written for human code. Here is how I test code an agent wrote so it holds up.EngineeringHow to Review an AI-Generated Pull RequestReviewing an AI-generated pull request needs a different checklist than reviewing a human's. Here is what to look for so bad code does not slip through.