Maestro Mission Control: One Person, One AI Factory, Every Phase
The hub of the Maestro series: Gopala's 18-month plan to take one self-improving (L4) idea to a frontier-style model and client products at start-up scale. See every phase and hat on one timeline, which layers of the stack he works on himself and which he delegates or buys, where the money goes, how the self-improvement loop is kept safe, and the risks that could end it. Everything links to the parts as they are published.
Open the timeline ↓ Start at Part 1 →New to AI? A 2-minute primer
Seven short animations explain everything the series assumes: what a token is, what a model really is, what a GPU does, how training works, why it takes 1,024 GPUs, how answering works, and how it all fits into an "AI factory". Each plays on its own; tap anything in a picture for a definition.
The whole system, live
Every layer of Gopala's AI factory in one picture. Pick a part of the series to see which layers it works on and how they interact at that moment: choosing (selection: comparing options), designing (deciding how the parts fit) or running (data and work flowing through). Or press play to watch the whole 18-month story.
1. The 18-month timeline
Phases overlap: racks are reserved while experiments run, evaluations are built before training starts. The red line is where the story is now. Tap a bar.
2. Hats × stack: what Gopala does himself
Rows are the layers of the live software stack and the 3D factory. ● hands-on · ◐ directs AI agents or contractors · ○ bought or rented · — not his job. Tap a cell.
3. The L4 loop and its safeguards
The self-improving loop at the heart of the plan (Part 1), with the fixed gate and tripwires that keep it at level 4. Press play; turn on a fault to see which safeguard catches it.
4. Money: plan versus spend
From Part 3's budget script and cash model: a $20M raise, released at evidence gates. Bars show planned spend per phase; the darker part is spent so far.
5. Risk register
Ratings by phase, updated as parts are published. Tap a cell for the mitigation.
6. All 18 parts
Built on: AI Compute Stack · Producing Intelligence · Who Runs the AI Factory · ML Layer · Data Layer · Evaluation · Agent Layer · RSI Evolution Map · Software Stack, Live · AI Factory in 3D.