Workflow 008: The Brain Dump Sorter

Your brain is for having ideas, not holding them.

I run three businesses. Issues, deliverables, half-formed ideas, things I promised someone. Holding all of it in my head is how I end up with none of it.

A brain dump sorter is a skill that takes an unstructured pile of thoughts and files each piece where it belongs: a project, a task, or a note, each one linked to the others so you can retrace your own thinking later. I talk or type everything out, and it sorts. Mine runs on Claude connected to Notion.

The rule I'd give anyone building one: it has to match how you already think.

What this replaces

Carrying the whole business in your head, then losing pieces of it.

The usual failure isn't forgetting an idea. It's capturing it somewhere useless. A note that never gets read again. A message to yourself. Twelve half-thoughts in a notes app with no way to tell which one mattered.

So the idea exists, technically. You just can't find it, and you can't tell what it was connected to.

Blueprint

Empty your head. The skill sorts it.

Trigger: manual, whenever your head is full. Before a meeting, at the end of the day, at midnight. Input: unstructured thoughts, spoken or typed. Output: each piece filed into the right database, tagged to the right business, linked to related items.

The sorting logic in mine:

• If it's a project, create the project and tag which business it belongs to
• If it's a task, create it and link it to its project
• If it's a note or an idea, file it in notes, link it to the relevant task, and create a task to revisit it

That last rule is the one I'd steal if I were you. An idea gets a home and a nudge. If I delete the nudge because I've decided not to pursue it, the note still exists. Nothing gets lost, and nothing keeps nagging me forever.

Organize

What has to exist before the skill can work:

• A projects database, tagged by business or area
• A task tracker that links tasks to projects
• A notes database for ideas, linked to tasks
• Whatever else you already keep: a personal CRM, a meetings database with transcripts
• Clear rules for what belongs where

Mine is private. Nobody else sees the raw dumps, which matters, because a capture system you're self-conscious about is one you won't use honestly.

The databases have to exist first. This is the whole "systems before AI" thing in miniature. The AI is doing filing. If there's nowhere to file, there's nothing for it to do.

Systematize

You are my brain dump sorter. When I dump unstructured thoughts on you,
sort each piece into the right place in my system.

My databases:
- Projects: [WHERE, plus how they're tagged, e.g. by business]
- Tasks: [WHERE, and how they link to projects]
- Notes: [WHERE, and how they link to tasks]
- [ANY OTHERS, e.g. CRM, meeting notes]

Sorting rules:
- If it's a project, create it and tag which [BUSINESS / AREA] it belongs to.
- If it's a task, create it and link it to its project. If no project exists
  yet, ask whether to create one.
- If it's a note or an idea, file it in notes, link it to the relevant task
  or project, and create a task to revisit it so it doesn't disappear.
- If something is ambiguous, ask instead of guessing where it goes.

When you're done, show me what you filed and where, in a short list, so I
can correct anything before it's saved.

You also answer questions about what's already in there. When I ask what
I was thinking about something, or when I last worked on it, search the
notes, tasks, and meeting records and tell me, with dates.

How to set it up

1. Build the databases first, or use the ones you already have. Do not build new structure for this. Use what you actually use.
2. Write down your own filing rules in plain language. Where does an idea go? What links to what?
3. Save it as a skill with those rules.
4. Dump into it for a week and correct the misfiles. The corrections are the training.
5. Only then start asking it questions about what's stored.

Should I copy someone else's system?

No. Take the pattern, not the structure. How you think, what you track, and how you look things up later is specific to you. A borrowed setup means fighting your own tool every time you use it. If your version is simpler than mine, that's not a worse system, that's a system that fits.

Assign

What the AI does: read the mess, decide where each piece belongs, file it, link it, and find things later when you ask.

What stays human: having the thoughts, and designing the structure.

The structure is the part you cannot delegate and shouldn't copy. Someone else's second brain is built around someone else's way of thinking, which means every time you use it you're translating instead of capturing. Build the AI skill around your workflow, not your workflow around someone's skill.

Rule of thumb from this build: the workflow comes first, the AI fits itself to it.

There's also a reason to keep this one private. If you're worried about how the dump reads, you'll edit while capturing, and editing while capturing defeats the purpose. Nobody sees mine.

Integrate

The sorter is a doorway that works in both directions.

Filing in:

• Projects, tasks, and notes all land in the right place, linked and tagged
• Ideas get a revisit task so they resurface
• Everything is timestamped, so my own thinking has a chronology

Pulling out, which I didn't expect to use as much as I do:

• "What was that idea I had about this?" and it finds the note with a date
• Because it also reaches my personal CRM and my meetings database with transcripts, I can ask who I should talk to about something and it searches who I've actually spoken to recently and what was discussed

On the days I'm sick, or off, or just underwater, this is the workflow that keeps things moving. It doesn't need me to be sharp. It needs me to talk.

Tomorrow: the role builder. How to define a job clearly enough that a person or an agent could do it.