Thoughtworks: don't ask AI to write user stories yet — ask it to find the decisions nobody made
rseroter · x · 2026-10-27
A Thoughtworks blog argues that pasting a vague request like "users should be able to export a filtered report" into an AI assistant yields a seemingly complete user story in seconds — but specific doesn't mean validated.
Does "filtered" mean every matching record or just the visible page? Which date field applies? Should hidden columns appear in the download? These are product decisions, not writing problems, and someone must make them. The core thesis: at the requirements stage, ask AI to find the decisions nobody has made yet — what you need is a list of questions for the team, not a finished story. AI is good at turning vague requests into confident-sounding language, and that confidence is precisely the trap.
More from coding & agent
- Cloudflare's Clairic: a System One extraction model returning typed values with confidence scores — omarsar0 · 2026-10-27
- Fanout: open-source tool tames parallel Claude Code and Codex sessions before merging — crumb_ona_breadstick · 2026-10-27
- Rust LVGL-style embedded UI library twine ships, built largely with Claude — Necessary-Pilot-3971 · 2026-10-27
- Share prompts, not code: a reusable prompt suite for evaluating small models (4B-36B) — yaroslavvb · 2026-10-27
- Maze Launches Maze Code: AI Agents That Investigate Dependency and Code Vulnerabilities — AndrewM_Webb · 2026-10-27
- rumi: a shared-doc workspace for orchestrating multiple coding agents at once — MuzafferMahi · 2026-10-27