I currently use 5 repos for my work.
- A repo for all my SEO, AEO and GEO planning and research
- A repo for my studio’s website
- A repo for my personal website
- A repo for my content pipeline
- A repo for my cold emails
Till now, these were isolated and working fine for what they were set up to do.
The challenge is that everything in my work is interconnected. The way a regular business should be.
My SEO strategy will impact my content pipeline and studio’s website. And in each repo, I’d do research and create plans, but some would be left midway or deprioritised.
There was no coherence.
So I thought of creating a ‘Company Brain’ like everybody else on Twitter.
This is a live draft of how I’m setting it up, how it’s going and what I’m planning to do next.
Version 1
I created a company-brain.yaml in the root of each repo, and it works as a map. This doc covers all the actions the agent can perform, approval gating, and data syncing.
Links to each of these docs are added to the new ‘company-brain’ repo with brief descriptions. It’s a rudimentary MCP-like approach to structuring access to the repos.
This works fine, but it’s obvious there are major drawbacks.
The maps and the documentation don’t do the work in themselves. It still depends on the agents to read, choose and follow the current path. The rules can drift across the repos and over time. And there is no decisions log or changelog. A Company Brain is supposed to store not just the future, but also past actions and the reasoning behind them.
Version 2
I decided to deepen the modules and add a small interface.
- brain ask — uses current decisions, open work, and source evidence.
- brain status — shows the active plans, blocked work, overdue actions, and things that need my decision.
- brain sync — read-only access for sub-repos and flags changes.
I also added a decision and action ledger with source linking and timestamps.
The parent repo decides the overall strategy, while sub-repos have a separate doc to track strategy changes, etc., so the parent repo can track them.
This is my current setup, and based on how things go, I might change or improve it.