Plan the software. Let the AI write it.

Practical guides to designing systems with an AI assistant: what to decide yourself, what to write down, and how to give Claude Code a plan it can read.

• How to use Hunch Board with Claude Code

  Run npx hunch-board init in your project. It writes board.hunch, a skill Claude Code can load, and a note in CLAUDE.md. Ask Claude to draft the board, correct it by hand in the browser, then ask it to build from the file and keep the file up to date.

  November 5, 2026 · 4 min read

• How to architect a software project with AI before you write code

  Don't ask the AI for an architecture. Have it interview you, draw a first version, and list what you haven't decided. Then make those decisions yourself, write down what's out of scope, go deeper only where it's risky, and build one slice from the plan.

  November 5, 2026 · 5 min read

• Why a plan helps AI write better code, and when it doesn't

  An assistant can read your code, but not the decisions behind it or what you want next. A short plan supplies that: what was decided, what's out of scope, and what done means. It won't replace tests or review, and we don't have controlled measurements of the gain, so judge it on your own project.

  November 5, 2026 · 4 min read

• CLAUDE.md or an architecture board: what goes where

  CLAUDE.md is loaded in every session, so keep it to commands, gotchas and rules that apply to every task. Repeatable procedures go in a skill. The design of the system and why it's that way go in a separate plan the assistant reads when the task needs it.

  November 5, 2026 · 3 min read