"The agent understood me" isn't a product requirement: put implicit decisions in the spec
shoumikgoswami · reddit · 2026-10-27
An AI product builder argues the more useful question than "can the agent build this?" is "what decisions am I leaving it to make?" Using the Elephant-Goldfish Model workflow—one agent writes a spec, fresh reviewers judge it without the original conversation—he shows how vague requirements like "keep sensitive data away from the model" hide real product decisions (speaker notes? text in images? uncertain detections? export logs?). A spec should answer these or mark them unresolved; a useful test is whether review exposes a missing decision that changes the spec, or just produces reassuring text.
More from coding & agent
- xAI releases free Grok Bot course: 18 real-team playbooks, ~3 hours, covering sales to engineering — victor_explore · 2026-10-27
- 100+ commits per day is the new bare minimum in the AI coding era — damnGruz · 2026-10-27
- Rocket launches open app platform to help vibe coders distribute AI apps — alexmacgregor__ · 2026-10-27
- Turn product photos into interactive 3D-style displays with AI video frames, no 3D modeling — aigclink · 2026-10-27
- Full walkthrough: interactive product displays from photos using GPT-6 Astra and AI video as UI — MaziyarPanahi · 2026-10-27
- DeepSeek leads four Qwen3.5-9B models to reverse engineer binaries — team mode costs more than solo — HauntingMoment · 2026-10-27