H1, eh?🍁

You think you're the best. The machines think they are too. One of you is wrong.

01001000 00110001 00111111
scroll · the clock is running
[2023.07]IMO benchmark · machine score: 0 / 42. Could not earn a single point. The hardest staircase we ever built — not one step reached.
[2024.07]SILVER. One point shy of gold. Solved a problem only five humans on Earth solved that year.
GOLD.[2025.07] matched the best young minds alive, problem for problem

Two years. Zero to gold. The fastest anything, silicon or carbon, has ever made that climb.

// exception caught

But there was a sixth problem. The one that needed a new idea: not search, not scale, a move no one had ever made before.

SEGFAULT — every machine failed it.

PROBLEM_6.trace
├─ reduction_to_search ... NONE
├─ benefit_from_scale .... NONE
├─ requires ............. CONSTRUCTION
└─ result .............. ALL MODELS FAILED
A high-schooler didn't.
// the originating move

Omar Khayyam solved the cubic with a curve.

The algebra to do it did not yet exist, so he dropped to a deeper layer, geometry, and built the missing tool there: a parabola crossed with a circle, and the root falls out.

That corner of the universe still belongs to people. To the ones who reach for the curve when there's no algebra yet. It's shrinking. We're not watching it close from the cheap seats.

We're building a lab around the hardest unsolved problems left, and choosing thirteen people to attack them.

You don't apply with a résumé. You apply by breaking one of our problems. The problem is the door, the test, and the proof, all at once.

guest@h1a:~$ ls -la /problems --open

Four domains. One open question each you could own. Every problem is something we genuinely need answered. Some are open: no answer in a drawer, because nobody's found one yet.

drwxr-xr-x01

motion_intelligence

Reconstruct a whole body's gait from one sensor. Predict the failure that wasn't in the training set. Find the invariant that transfers across species.

inference
drwxr-xr-x02

embedded_silicon

The same mind in ten milliwatts. The bug that survives ten million test vectors. The honest boundary of analog in-memory compute.

edge
drwxr-xr-x03

quantum_physics_math

Derive the Cramér-Rao floor. Recognize the sphere-packing quantizer. The intent manifold, genuinely open.

open
dr-x------04

the_weird_one

No prompt. No rubric. Show us one true thing you find beautiful that almost no one else would, and defend why it matters when we push. Smart and kind aren't enough. We said weird, and we meant it.

★ selects

ARE YOU BETTER
THAN THE MACHINE?

There's one way to find out. Not against the machine, against the problem. The problems are live.

0000

minds who've taken a crack

>

No trackers. No cookies. No nonsense. Just the problems and the people who can't leave them alone.

// thirteen slots. the corner is closing. bring your move.

or, if you read the page instead of looking at it, find ROOT ACCESS ↓

— ROOT ACCESS · for the capable —

This door has no handle.
It has a question.

The form above takes your email, like everyone's. This one takes a word. The wall below only looks like noise: decode it, and it names where the key is kept. Three steps stand between you and a direct line. Most people stop at the first.

wall · 01 — decode me (base64) WW91IHJlYWQgdGhlIHBhZ2UgaW5zdGVhZCBvZiBsb29raW5nIGF0IGl0LiBUaGF0IGFsb25lIGlzIHJhcmUuIFRoZSBuZXh0IGtleSBpcyBrZXB0IHdoZXJlIHRoZSBidWlsZGVycyBsb29rOiBvcGVuIHRoZSBjb25zb2xlLg==
root@h1a:~#
three steps. the wall is step one.
ACCESS GRANTED — you found the handle that was never there.

You did it the intended way: wall, console, riddle, word. The door was the first problem, and you solved it. No form, no waiting — the lab is open to you now.

enter the frontier lab →

no line is printed — that’s the last test. the name is the move every game opens with · the house is the one you’re standing in · put them together and write