Coding agents can produce more software while leaving developers with harder decisions. That is the useful tension in Simon Willison's 2026 retrospective: he describes agents taking over easier implementation work while goals, constraints and judgments about the result still demand his attentionThe mechanism that lets a model weigh which earlier words matter for the word it's currently processing — the core operation of a transformer.Full definition →.
His experiments with generated games make the distinction concrete. He could obtain something that looked and behaved like a game, yet found the results entertaining only briefly. Producing working mechanics had not supplied compelling gameplay. His account suggests that faster implementation increases the importance of deciding what deserves to be built and recognizing whether the result serves its purpose.
Terms in this piece · Glossary
attention — The mechanism that lets a model weigh which earlier words matter for the word it's currently processing — the core operation of a transformer.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
