Quick fixes AG-W002

Stop re-explaining your project

Shared methods · A shared method; linked tool guides explain the exact steps.

Typing the same background into every chat is a tax. Put the stable facts somewhere persistent once, and every conversation starts warm.

Quick fix: Stop re-explaining your project

Instead of Start every chat with the same three paragraphs about your project.

Try Save a short project brief. Reuse it in the next conversation, or put it in your tool's project instructions when that feature is available.

Write the context once, in the place built for it.

Why this works

Do not assume a new conversation has your previous context. A short, maintained brief gives you a reusable starting point. Check what your tool actually loads before relying on persistent instructions or memory. The written version is easier to correct than background retyped from memory.

When this applies

You catch yourself typing "for context, I'm working on…" for the third time this week, or pasting the same document into every new chat.

Template

What goes in the persistent version:

What the project is, in two sentences. The constraints that never change (audience, format, tech, deadline). What good output looks like — ideally one example. What to never do.

Leave out anything that changes weekly; stale context misleads more than no context.

Go deeper

For OpenAI / Codex, see project instructions and skills. The following guides cover Claude specifically.

Projects: what belongs in one, what poisons it is the main tool here, with memory across chats for durable personal facts. For code work, a CLAUDE.md that gets followed is the same move, and a prompt you re-use verbatim can become a slash command.

Improve next

Last verified · Reviewed by Timothy Fehr