Every bot detection system in production today tries to distinguish humans from machines by analyzing what they produce — mouse paths, browser fingerprints, behavioral patterns. All of these can be synthesized. Origin Standard doesn’t detect bots. It measures the thermodynamic cost of being present on real hardware with a real nervous system. When faking presence costs the same as actual presence, the bot loses its only advantage: scale.
The Orthogonal Triad
Φ
Temporal
Sequential VDF proof that cannot be parallelized. The computation takes real time on real silicon. Fold points capture behavioral state at specific hash positions. You paid the cost, or you didn’t.
Ψ
Entropy
Fractal analysis of timing jitter from the browser’s event loop. Human nervous systems produce stochastic noise with specific DFA characteristics. Computed values have different spectral signatures. Biology is messy. Math is clean. We measure the mess.
Ω
Environment
Cross-correlation of hardware signals — timer precision, event loop variance, thermodynamic skew, oscillator drift. Real hardware hums. Virtual hardware is silent. The gate listens for the hum.
The Invariant
Score = Φ × Ψ × Ω
Multiplicative. Not additive. If any domain is zero, the product is zero. There is no gradient to climb. There is no partial credit. The bot either produces real signals across all three orthogonal domains simultaneously, or it scores nothing. The oracle is dead — PASS and FAIL are identical on the wire.
Tested. Not Theorized.
265+
Adversarial Attacks
0
TAOs Acquired by Bots
6
Attack Classes Tested
0.6+
Human Health Score
How It Works
01
Drop in. Not bolt on.
One JavaScript file. Your existing Node.js infrastructure. No onboarding, no training, no downtime. The gate runs alongside your content — invisible to the user, impenetrable to the script.
02
The gate fires in the background.
While the human reads your page, the gate silently collects hardware entropy, solves a sequential VDF, measures environmental signals, and submits a sealed proof. The human never notices. The bot can’t produce what it doesn’t have.
03
Three blind evaluators. Unanimous or nothing.
Φ, Ψ, and Ω evaluate independently. They don’t see each other’s results. All three must pass. The multiplicative model ensures that a perfect score in two domains with zero in the third produces zero. No partial credit. No gradient for the attacker to optimize against.
04
The oracle is dead.
PASS and FAIL responses are identical on the wire — same size, same timing, same structure. The attacker receives zero feedback. No gradient. No hill to climb. Every failed attempt doubles the compute cost of the next one. The economics kill the attack before the physics do.