Convert a spoken brief into a working task tree

A developer fed a spoken project brief into LLM agents, which turned it into a spec, plan, scaffold, module tickets, and PRs.

Done with Custom agent· Advanced· Oct 2026

What they did
The developer began with a messy spoken outline of a project. Whisper turned that recording into text, and an LLM then compressed it into a markdown spec that was saved as issue 1. After the developer edited it and a second LLM pass expanded it again, another LLM mapped out the build and created the starter scaffold. An agent then split the work into module tickets, other agents handled commit-sized tasks as PRs, and a critical reviewer plus follow-up agents kept iterating.
How it went
The author says the workflow produced strong results. Most PRs were accepted unchanged, and when an agent chose an off-track solution, it was corrected during the later passes. They also said it cost far less than hiring human developers, even with many agents and CI runs.
Tips
This works best when the codebase is broken into modules with little overlap between them. The shared workflow, style, and lint rules live in one repo and get vendored into each project to keep agents consistent.

Try it yourself with Claude Code

I have a rough spoken brief for [project idea]. Turn it into a clean spec, list any open questions, then draft a module-by-module implementation plan. Create the initial scaffold in [language/framework] and write one markdown ticket per module in a [tickets] directory, each with scope, acceptance criteria, and dependencies. Stop after the scaffold is running and the tickets are written, then show me everything for approval before you do any sending, deleting, submitting, or other irreversible action.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on news.ycombinator.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.