The operator
You can’t buy AI transformation. You install it — one human, one role at a time. Owned by them.
Michael Sebastian · Founder, Branded Mayhem Collective
As featured in
The work
I run a brand studio. Branded Mayhem Collective — small team, founder-direct work, the kind of client roster you build by saying no to the wrong fits. In 2023 I started installing AI into the actual work, not as a tool laid on top of it. Three years later, the install layer is more valuable than any single deliverable the studio has ever shipped.
My corpus is two-and-a-half million tokens of decision history, brand frameworks, client context, methodology. It runs as an MCP server gated to a branded subdomain. It answers from Claude, ChatGPT, Cursor, any LLM that speaks the protocol. When the model changes — and it will — the install moves with me. The platform is the substrate. The install is the work.
M.L. Sebastian is what I built so other operators can have the same thing. Same architecture, same portability, same respect-for-the-work posture. The unit is the person.
The anchor
In The Measure of a Man— Star Trek: TNG, season two — Starfleet’s position is that Data, the android, is property. A piece of equipment owned by the fleet that built him. Captain Picard argues otherwise. Not that Data is sentient in some abstract philosophical sense — that’s a fight you can’t win on the witness stand. He argues something narrower and sharper: that the question of who owns the intelligence determines what kind of culture you’re going to build with it. The verdict was that Data could choose.
I think about that scene a lot. The corporate AI buyout of the next decade is the inverse of that argument — the working intelligence you accumulate at your job belongs by default to the platform you accumulated it on. Your context, your patterns, your memory. You don’t get to choose. The leverage runs one direction.
The install I build runs the other direction. The brain belongs to the person who owns the role. Same architecture as the vendors’ lock-in version. Different owner.
The moat
In the first era of AI, the moat was the model. Benchmarks, context windows, training runs. That race is mostly over — the margins between frontier models have compressed to where it’s no longer the primary axis of competition.
In the era we’re entering, the moat is memory. The always-on agent that knows you because it’s been watching. The platforms that hold your accumulated working intelligence — your domain encoding, your workflow patterns, your behavioral calibration — have built the deepest lock-in the technology industry has ever produced. Not because the model is better. Because the switching cost of abandoning six months of compounding context is unthinkable.
Your AI working intelligence is the most valuable professional asset you’re building this year. It belongs to the platforms. Not to you.
I built M.L. Sebastian to change that. One human, one role, one Personal Brain — owned by the person, portable when they leave. The vendor’s platform is the substrate. The install is the work.
Pick your fighter carefully. Or build your own.
Two days ago at Meta
On May 20, Meta started laying off about 8,000 employees. Roughly 10% of the company. The reason given is the same as every recent round of tech layoffs — a pivot to AI. In a leaked all-hands recording, Mark Zuckerberg told his employees that their devices are being tracked to train Meta’s AI models. His reasoning: “the average intelligence of the people who are at this company is significantly higher than the average set of people that you can get to do tasks if you’re working through these contractors.” The workers being laid off — and the ones still there — are training the systems that replace them. [source]
The week before, Marc Andreessen went on 20VC and called the labor-displacement narrative “100% incorrect” and a “classic zero-sum fallacy.” His framing: companies were overstaffed by 50–75% after COVID and AI is the convenient excuse for cuts that were going to happen anyway. [source]
Both things can be true at once. Companies were overstaffed AND AI is the mechanism — and the cover story — for the cut. Either way, the worker loses. The pattern is durable: capture the worker’s knowledge first, then remove the worker.
What you’ve been training without knowing it
Every time you write to ChatGPT, refine a Claude project, accept a Copilot suggestion, or let a Gemini transcript get saved — you’re teaching that platform how YOU work. Your vocabulary. Your decision patterns. Your tolerance for risk. The shape of your judgment.
That’s the most valuable training data the industry has, because it’s the working intelligence the company actually paid for. By default, that intelligence belongs to the platform. Or to the company, through the platform. Not to you.
The fingerprint that makes your work yours — the years of judgment that made you the person they hired in the first place — gets harvested into a training set you didn’t consent to and can’t take with you when you leave.
The counter-move
The Personal Brain Install is the same architecture as the corporate version — files in, brain out, retrievable from any AI tool. The difference is the owner. Your brain lives on a subdomain you brand. The data is yours. The recipes are yours. The retrieval contract is shaped to YOUR role, not a job-family average.
When you change jobs, your brain comes with you. When the company you work for gets cut to AI, your asset is the brain you built — not just the final paycheck. When the next model launches, the install adapts to it instead of starting over.
Plug your brain into the work. Let it share in the value. Don’t hand it to a platform racing to do without you.
The rules
01 · The brain
Not to me. Not to the platform. Files in, brain out — gated to a subdomain you brand, walk away with the day you stop working with me. Or the day you leave the job. The portability is the product.
M.L. Sebastian is the AI consultancy arm of the Branded Mayhem Collective.
