Lobe 05 · Wire It In

Wire your brain into your agent

Time to make it real. Hand your brain (your folder of knowledge files) to Claude or ChatGPT, then ask it something only your brain knows — and watch it answer without you re-explaining a thing. That moment is the whole point of all this.

1Give the brain to your agent

Your brain is just a folder of plain text files (the .md ending means "markdown" — ordinary text any AI can read). Because it's plain text, you don't need anything fancy: you just hand the files to a normal chat. Pick the tool you use — the free version works perfectly. Here's the exact way to do it in each.

Free Claude

The free, no-setup way

In a browser, go to claude.ai and start a new chat. Then:

  • Look at the bottom of the message box for the paperclip icon (it may also look like a ＋ or say "Upload"). Click it.
  • Choose Upload from computer, find your brain folder, and select all the .md files inside it (hold Ctrl, or ⌘ on a Mac, to click more than one).
  • Wait for the little file chips to appear above the box, then send your first message.

No upload button, or it won't take a file? Open each .md file, copy all the text, and paste it straight into the chat. Do this for every file. It's less tidy, but it works exactly the same.

Free ChatGPT

The free, no-setup way

In a browser, go to chatgpt.com and start a new chat. Then:

  • At the bottom of the message box, click the paperclip or ＋ button, then choose Upload from computer.
  • Find your brain folder and select all the .md files inside it (hold Ctrl, or ⌘ on a Mac, to pick more than one).
  • Wait for the files to finish attaching, then send your first message.

No upload button, or it won't take a file? Open each .md file, copy all the text, and paste it straight into the chat — one file after another. Works the same way.

The same brain works in both tools — you don't rewrite anything. So once it works in one, try the exact same files in the other. That's the whole idea: write your knowledge once, and any AI can read it.

Already comfortable with a code editor or the command line? You can also keep the brain in a project folder and let a coding tool (like Claude Code or the ChatGPT/Codex command-line app) read the files where they sit — point it at index.md first. If that sentence meant nothing to you, ignore it completely: the upload steps above are all you need.

2Test it with a "brain-only" question

Now ask a question it could only get right if it actually read your files. In that same chat (right after the files are attached), copy this and paste it as your message — type it exactly:

Using the files I just gave you, what am I working on right
now, and which tools do I use? Answer in my voice.

If it answers correctly without you explaining anything first — your brain works. Congratulations, you just handed an AI your knowledge.

If it gets something wrong, don't blame yourself or the AI. The fix is almost always a clearer file — for example, the answer lives in a file but it's buried or vaguely worded. Open that file, make the wording plainer, save it, then upload again and re-ask. (A "concept file" is just one of your knowledge files — like tools.md or voice.md.)

3Link your files to each other

This turns a pile of separate files into a connected brain, where one file can point to another. You do it by adding a simple link inside your text. The pattern is square brackets around the words you want to show, then round brackets around the file name: [words to show](file-name.md). For example, open projects.md and add lines like these:

I'm building my project using my usual stack — see [tools](tools.md).
My writing style for it is in [voice](voice.md).

Now, when the AI reads one file and sees a link, it knows to go look at the related file too — instead of treating every file as a stranger. (If you don't add these links, that's fine — your brain still works. This just makes it tidier and helps the AI connect the dots.)

✅ Lobe 05 complete when: your AI answers a brain-only question correctly — ideally in both Claude and ChatGPT — and (optionally) your files link to each other. Next up — Lobe 06: ship it and enter the comp.

Tip: keep testing with real questions you'd actually ask. Every wrong answer is just a file that needs to be clearer — that's you improving your brain.