Sensitive Dependence on Initial Conditions
·Mentoring·
The Mathematics of Machine Intelligence · Age Six to the Frontier
σ = 10·ρ = 28·β = 8/3·Δt = 0.004
Two trajectories · one ten-millionth apart at the start · watch the gold leave the blue
School mathematics is a museum — arithmetic performed as ritual, formulas kept behind glass, and nobody is ever told what any of it is for. I teach the mathematics that machine learning is actually made of. And I start at six, with counting — because counting, done honestly, is already the first lesson in structure.
From counting to the research frontier. One road, no gates. The endpoint is not a grade — it is a child who opens this week's listings on arXiv, reads the papers, reproduces them, and writes the code: an ML researcher and an ML engineer in one head. But — and this is the whole of the small print — not for everyone.
A mind is not a vessel to be filled. It is a circuit to be closed. Once it closes, it oscillates on its own, and my work is done.
Initial Conditions
Everything begins with a measurement. Before the first lesson I find where the child actually stands — not the school grade, the true coordinates. Lorenz proved that a millimetre at the start becomes a continent at the end. I place that millimetre deliberately — and the earlier it is placed, the further it travels.
Resonance
I do not push. Pushing is what schools do, and the amplitude stays at zero. I find the natural frequency of one particular mind and drive it in phase — small forces, perfectly timed, until the swing is enormous. Tesla lit cities on this principle.
Self‑Excitation
The only thing truly worth teaching is how to learn alone. Everything I know I taught myself first and was examined on second. The child leaves with the generator, not with a jar of sparks.
I · Nvmervscounting · age six to eight
Counting, place value, arithmetic as structure — never as ritual. Patterns hunted and named. A function met for the first time as a machine: a box that turns 3 into 6, and the child asks what is inside the box.
II · Strvctvraalgebra & functions · eight to eleven
Letters as numbers not yet chosen. Equations, coordinates, graphs — and the first Python: the computer as an obedient, pedantic pupil who does exactly what was said, never what was meant.
III · Spativmvectors & data · eleven to thirteen
Vectors, and matrices as motions of space. Probability learned by running the experiment ten thousand times. Real datasets on the table long before the word “statistics” is ever said aloud.
IV · Motvscalculus & linear algebra · thirteen to fifteen
The derivative as sensitivity; the gradient as the direction downhill. High-dimensional space entered without fear. Probability again — this time with proofs.
V · Machinathe mathematics of learning · fifteen to seventeen
Loss surfaces, gradient descent, backpropagation as the chain rule wearing overalls. Information theory. Networks built bare-handed in NumPy, then properly in PyTorch.
VI · Fronsthe frontier · seventeen, or when ready
Read this week’s listings, reproduce the papers, write the code, find the gap. The road ends where the field currently does — and then it doesn’t. arxiv.org/list/stat.ML/recent →
This is for the child who asks why twice. Who takes the toy apart and is happier with the gears than with the toy. Age and current grades are irrelevant; the second why is not. And six years old is not too early — it is exactly on time.
It is not for everyone. Not for the parent buying grades by the hour. Not for a child who would sincerely rather be elsewhere — a circuit with no natural frequency cannot be driven, and I will not pretend otherwise. Admission is by application. No prices are published; terms are disclosed on acceptance.
- Electrical engineer
- Computer scientist
- AI researcher & engineer
- Applied-mathematics scientist
- Self-taught first · examined second
Admission is by letter. Write what has been observed: the child, the age, the second why. If the initial conditions are right, you will hear the resonance.
One letter, in your own words — the child, the age, what has been observed, and where to answer.
SCRIBE LIBELLVM →the direct wire is cut — the letter travels by the petition
ANSWERED PERSONALLY · NOTHING IS AUTOMATED
No prices are posted here. Terms are disclosed on acceptance — and when they are, payment travels one of three wires. The first is the default.
WIRE I · BANK TRANSFER · FASTER PAYMENTS · THE DEFAULT
ACCOUNT NAME
Stanislav Sobolev SORT CODE
30-97-81 ACCOUNT NUMBER
11181862 GBP · ARRIVES IN SECONDS
WIRE II · LLOYDS · REQUEST A PAYMENT (LINK PAY)
pay.lloydsbank.co.uk/pay/?a=SSobo-VTSZB2 SCAN IN YOUR BANKING APP · OR OPEN THE LINK
WIRE III · CRYPTOCURRENCY · USDT
VIA TRON · TRC-20
TB7fAfRTxBCg3UEVuGP6C9RvqjgswgFP7M VIA ETHEREVM · ERC-20
0x92311a146c0b3dbb10ebdb2735c4c48de422a130