Your Brain Shouldn't Be SaaS

Someone is going to sell enterprises a Brain. The better answer is already sitting in plain markdown on your own machine.

Last updated April 17, 2026

The Tweet That Captures the Moment

Alex Lieberman (@businessbarista) recently wrote that someone is going to build a world-class "Brain" for enterprises and make a stupid amount of money. He quoted @da_fant:

Coding with AI is solved because all context is in the git repo. Knowledge work is difficult because context is spread out. An AI system that creates a git repo with all context for a knowledge worker will be able to 100% automate the work.

That is the clearest external articulation of the thesis I have been building toward for two years. The git repo is the secret. Engineers have been ready for AI because their context was already centralized, versioned, and plain text. Everyone else is paying the price of having their context trapped in Slack, Notion, Hubspot, Granola, Gmail, and eighteen other SaaS silos.

Lieberman is right about the problem. He is wrong about the solution.

The Three Real Barriers

Lieberman names three reasons knowledge work has not been AI-ready. He is exactly correct on the taxonomy:

  1. Distributed. Your context lives in twenty tools. Transcripts in one, docs in another, customer data in a third, Slack threads nowhere you can search, SOPs in whatever random Google Drive folder someone made in 2023.
  2. Unstructured. Even if you ingest it, it is a mess. There is no schema. There is no versioning. There is no clean way to pull "the specific asks from the sales call, the previous proposal format, and the completed sprint board" into one coherent document.
  3. Unverifiable. Code either passes the test or it does not. Knowledge work has no unit test. Is this proposal good? Is this content in your voice? Is this idea novel or slop? The system has to improve on judgment, not correctness.

This taxonomy is a gift. Take it. But the conclusion Lieberman draws ("someone will build the Brain and sell it to enterprises") is the wrong fork.

The Fork

There are two answers to the three barriers. Only one of them ends well for you.

Answer A: Rent your Brain. A vendor builds a centralized system that connects to your SaaS sprawl, ingests your context, maintains your schema, compacts your memory, and sells it back to you as a subscription. Your entire operational intelligence lives inside their product. You are now dependent on their uptime, their pricing, their roadmap, and their willingness to export your data when you want to leave.

Answer B: Own your Brain. You build a markdown-first repository of everything that matters about your life or your business. You commit it to git. You give your AI agents read and write access. You run the sifting, the organization, and the compaction yourself, with AI doing the heavy lifting. The Brain is a folder. You can zip it up and move it tomorrow.

Answer A is the obvious VC bet. Someone will make a lot of money building it. That does not mean you should buy it.

Why SaaS Brains Are a Trap

A Brain is not a productivity tool. It is the substrate of every decision you make. Once it lives inside a vendor's walls:

  • Exit cost becomes terminal. Leaving Notion is annoying. Leaving your Brain is impossible. Every memo, every pattern, every pricing model, every piece of founder judgment is now shaped by a structure you do not own.
  • Your context trains their model. The value you created becomes the moat they use to lock in the next customer. You are the raw material.
  • Lock-in compounds. Each month you add context, the switching cost grows. By year three, you have poured so much of yourself into their silo that alternatives do not feel real.
  • The AI you get is not yours. A vendor-shaped Brain produces vendor-shaped outputs. Your agents start to think the way the platform wants them to think.

This is what I mean when I say the lock-in is coming. Centralized Brains are the final form of capture. They are Gmail for your thinking, Hubspot for your judgment, Salesforce for your strategy. You do not want that future.

The Sovereign Answer

The git repo insight is correct. The response is not a new SaaS. The response is to realize the git repo is already sitting on your laptop, and you can fill it yourself.

Plain markdown files in a folder, versioned with git, readable by any AI you choose. That is your Brain. Andrej Karpathy, one of the most respected researchers alive, keeps pointing at exactly this pattern: indexing documents, compiling a wiki of markdown files, querying it, filing results back in. No vendor. No subscription. No lock-in. Just files.

The distributed problem becomes an ingestion problem you solve once. Pull from your sources, drop into markdown, commit.

The unstructured problem becomes a schema you design for your own life. Folders are fine. Headers are fine. Clean prose is fine. You do not need a proprietary memory architecture. You need an opinion about how your work is organized.

The unverifiable problem becomes a feedback loop you run yourself. You read what the AI produced. You correct it. You commit the corrections back. Your Brain sharpens with every pass. This is reflexive self-improvement, and it only works when you own the substrate.

The Operator Version

If you run a small business or a team, the same rule applies. Your Sovereign Agentic Business OS is a git repository full of markdown. Principles, strategies, playbooks, client files, SOPs, meeting notes. Your agents read the repo. Your team reads the repo. When a new hire joins, they clone the repo and absorb years of context in a week. When you fire a vendor, nothing breaks.

This is the sovereignty stack applied to the substrate of decision-making. The moat you build inside a SaaS Brain is rented. The moat you build inside your own repo is yours forever.

The Timing

Lieberman is right that the market is about to wake up to this problem. He is right that someone will make a lot of money. But the people who bet on themselves, who build the sovereign version in plain text, are the ones who end this decade with leverage instead of dependence.

The tools are here. Claude Code. Cursor. Local models. MCP servers. Every piece of infrastructure you need to run a first-class Brain on your own machine is shipping right now. The only thing you have to do is start writing it down. Externalize your brain into files you control.

Do not wait for a vendor to sell you back your own context. Do not pour your judgment into a silo. Do not rent what you can own.

The enterprise Brain market is real. The sovereign Brain is the better answer. Someone is going to make a shit ton of money selling centralized Brains to companies that did not know they had another choice. You have another choice. Build it in plain text. Commit it to git. Own your context forever.