Skip to content

The House-Sitter Note

The Two Ways to Brief a House-Sitter

Imagine you're going on a two-week vacation. One friend is house-sitting for the first week, and a different friend is covering the second week. Neither has been to your house before. You have two options:

Option A: Call friend one when they arrive and explain everything — where the coffee is, the alarm code, when the dog gets walked, that the neighbor's cat might come by and shouldn't be let in. Then, when friend two arrives for week two, call them and explain the exact same things all over again.

Option B: Leave a note on the kitchen counter. Both friends read the same note when they arrive. Same information, no repeated phone calls.

That's exactly what's been happening with AI. Every new conversation is a new house-sitter showing up — and you've been making the phone call every time. Option A works, but it's exhausting and wastes time you could spend building.

Option B is what you're about to set up. A project context file is a note your AI coding assistant reads automatically at the start of every conversation. You write it once. Every new conversation reads it. No more repeating yourself.

What This Changes

Without a project context file, every conversation starts like this:

This is an Avalanche Field Guide for the Wasatch Range. We're using live avalanche forecast data and weather data. The main screen is the one you created last time. We want it to look like a professional safety resource...

With a project context file, every conversation starts like this:

As a backcountry skier, I want to see wind speed in the weather section so that I can assess whether ridgeline travel is safe.

That's it. AI already knows the project, the data sources, and the design. You go straight to building.

The Table of Contents, Not the Whole Book

Here's the thing about that house-sitter note: you wouldn't write a 50-page manual about every object in your house. You'd write a short note covering the essentials, and maybe mention where to find the fuller instructions — "the binder in the kitchen drawer has more information on the security system, the vet's number, and the weekly trash schedule."

Your project context file works the same way. It's a table of contents, not the whole book. It tells AI:

  • What the project is (one sentence)
  • What the main parts of the application are
  • Where to find more details (like existing documentation)
  • Any important rules to follow

If your project already has documentation, your context file should point to it, not copy it. This keeps the context file short, which matters because of the oxygen tank: everything in the context file takes up space in AI's working memory. A short note that points to good documentation beats a long one that duplicates everything. There's another reason to keep it short: AI pays the most attention to what's near the beginning and end of what it reads, and tends to skim over what's in the middle. A long context file doesn't just use up space — the points buried in the middle are the ones most likely to be ignored.

Team Activity: Write the House-Sitter Note

Format: Mob Session Time: ~3 minutes Setup: Gather around one screen. Open a scratch notepad or notes app — not the AI coding assistant. You're drafting first.

Pretend your AI coding assistant is the house-sitter arriving at your project for the first time. As a team, write a short note (5–8 bullet points) answering:

  • What is this project? (one sentence)
  • What are the main parts of the application and what does each part do?
  • What type of informational data does the project use?
  • Are there any rules about the domain that should always be followed? (Hint: "always use the official North American Avalanche Danger Scale colors — green, yellow, orange, red, and black — for danger levels 1 through 5")
  • Are there any rules for how the AI assistant should behave? (Hint: "always save and sync when I ask," or "challenge me when I don't provide clear acceptance criteria")

Keep it short — remember, table of contents, not the whole book.

Hold onto this list. You'll turn it into a real project context file in the next section.

Key Insight

A project context file solves the biggest frustration of working with AI: repeating yourself every session. Write it once, AI reads it every time — automatically. Keep it short and point to your documentation instead of copying it. The less you put in the file, the more room AI has for the actual work.