← Themes·NordenUse this theme

Merging agents, compute and rival review into true discovery.

On the record:

BigBounceDESI DR1ZenodoarXiv.PreprintsReviewed / 5xhubify.app

Houston is committed to a tomorrow where anyone with a question worth answering can run a real laboratory.

In pursuit of this purpose, in 2026 he founded Hubify Labs, a platform backing self-taught researchers with agents, compute, and adversarial review from five rival AI labs.

Hubify also runs BigBounce, a six-paper cosmology program — 17.65 million spectra analyzed in one weekend for $200 — as living, publicly reviewed proof.

The method

A laboratory in four movements.

01

Ask

Type a research question in plain English. That's the whole interface — no grant, no lab affiliation, no permission slip.

02

Run

Your agents pull real survey data, spin up GPU pods, and run the experiments around the clock while you steer.

03

Survive

Reviewers from five rival AI labs attack every claim. The kills are published next to the passes, round by round.

04

Publish

What survives ships as a living paper with its full review record — data, figures, and the failures too.

Why rivals

Cooperative agents flatter you. Rivals don’t.

Agent teams that share a goal converge on each other's mistakes. Hubify hands every claim to five rival AI labs with nothing in common except the incentive to break it — and a blind integrity audit holds any run whose verdicts look engineered. When a result survives, you can defend it anywhere. It already survived worse.

AnthropicOpenAIGooglexAIPerplexity

On the record · BigBounce

17.65M

DESI spectra, one weekend, $200

268,519+

Validated anomalies, four surveys

8.47M

Galaxy chirality catalog

6

Papers at the submission gate

Every round is public — read the review record, kills included.

Bring a question worth answering.

The agents are idle. The reviewers are waiting. The record is public.

Request access