The Third Brain · Coming soon
Stop re-explaining yourself to Claude at the start of every session
This course sets up Claude so it starts each session already knowing how you work. You put rules in instruction files, facts in memory, and repeat jobs in skills, then prune all three once a month. The fix is less text in the right place, not a longer CLAUDE.md that Claude stops following.
The course: Memory for your AI: instructions, skills and memory that last
$19 $39 early-bird price
This course is being built now. Join the early-access list to reserve the early-bird price and shape what it covers.
Joining the list is free. You only pay if you buy at launch, at the early-bird price.
If this sounds like your week
- Every new session you re-explain the project, your preferences and the decision you made last week.
- You wrote a rule down once, and Claude followed it for a day and then drifted back.
- Your CLAUDE.md has grown by accretion, and old instructions now contradict new ones.
- You are not sure what belongs in instructions, what belongs in memory, and what should be a skill.
What you do now, and where it stops working
Today you paste a context block into each chat, or keep adding lines to one instruction file every time Claude gets something wrong. The file gets longer, the rule you care about gets buried, and you are back to correcting the same things.
What changes after the course
- A new session starts with your role, your rules and your project facts already loaded.
- Your instruction files stay short enough that you can read them in a few minutes.
- Jobs you repeat each week run from a skill you wrote once, not a prompt you retype.
- Once a month you remove what is stale, so old decisions stop overriding new ones.
How it works: Three places, one prune
- 1Instructions: a personal ~/.claude/CLAUDE.md for how you work and one per project for that project, each kept to rules that apply to most tasks.
- 2Memory: Claude Code keeps an auto-memory index, MEMORY.md, and loads its first 200 lines each session, so the course shows you how to keep that index short and accurate.
- 3Skills: each repeat job becomes a SKILL.md in .claude/skills, loaded only when the job comes up.
- 4Prune: a monthly review, itself a skill, that finds contradictions and stale entries and removes them.
What you get
- Instruction templates for you and for each project
- Skill templates with worked examples
- Memory review checklist and skill
- Video walkthrough of a working setup
This is for you if
- You use Claude Code every week and repeat yourself more than you would like.
- Your instruction file has grown and Claude follows it less than it used to.
- You want a setup you can hand to a teammate and have it behave the same way.
Skip it if
- You only use Claude in a browser and do not plan to try Claude Code; the course is built on Claude Code files.
- You want a plugin that remembers everything automatically; this course keeps memory small on purpose.
- You open Claude a few times a month; the setup will not repay the hour it takes.
Before you join
- I already have a CLAUDE.md and Claude ignores it. Why would this be different?
- A long file mixes rules that apply to every task with notes that apply to one, and the rule you need gets lost. The course moves each line to the place it belongs, so the instruction file holds only what applies to most of your work.
- Will memory fill up with junk and make answers worse?
- It can, which is why the course treats memory as an index you review, not a log that grows. The monthly prune removes stale and contradicting entries before they pile up.
- I use Claude in the browser. Is this for me?
- Only if you are ready to start using Claude Code. The instruction files, memory index and skills the course builds are Claude Code features, and the first module walks you through installing it.
- Is this not the same as the free starter kit?
- The free kit gives you a CLAUDE.md and memory setup to copy. The course explains why each part goes where it does, adds skills and the monthly review, and shows you how to adapt the setup as your work changes. Jay West wrote and tested the kit on Claude Code and runs his own work on the full system.
- Do I need to be a programmer?
- No. Instruction files, memory and skills are plain-text files written in ordinary sentences. If you can edit a document, you can keep this setup.
Rather have it built for you? Describe the project and the same work can be done for your team.