Most quant edge is fake. We prove the few that are real.

A systematic manager claims a track record, a Sharpe, a return, a drawdown. The allocator cannot check it. The manager will not reveal the strategy that produced it, and rightly so: the strategy is the entire asset. So capital is deployed on trust: on a PDF, a reference, a brand.

That trust is routinely misplaced. Backtests are inflated, overfit, survivorship-biased, or quietly lookahead-contaminated. The allocator finds out after the drawdown. The honest manager, meanwhile, has no way to prove they are not one of the fakes.

The market runs on a claim nobody can check. We made the claim checkable — the one product every quant told "just show me the strategy," and every allocator who ever signed on a PDF, has quietly wished existed. It could not be built until now, because it needed two fields that never spoke: the statistics of overfitting, and zero-knowledge cryptography.

MIZAN re-runs the real backtest inside a zero-knowledge proof. It emits a credential that anyone can verify — that the strategy performed as claimed, after real costs, with no lookahead and on bound, point-in-time data — while the rules, parameters, and source stay provably hidden. The same proof now extends to AI itself: a neural network’s entire track record verified while its weights stay sealed — the model runs inside the proof, and is never revealed to anyone. The strategy never leaves the prover's machine.

• The manager gets a portable proof of a real edge, without exposing the alpha.
• The allocator gets a verdict they can re-check themselves, trusting no one, not even MIZAN.
• The verdict is binary and honest: a losing strategy — or one that survives only in-sample — earns no credential. The gate refusing a bad strategy is the product.

That panel is the live product, the engine at studio.mizan.market, where you can watch a credential get minted and re-verified in seconds. Every credential carries a deflated-Sharpe significance test — Bailey and López de Prado's own correction for backtest overfitting and multiple testing — held to the honest 95% bar.

On MIZAN's own flagship it reads not significant — and this month it refused our own AI: a neural network that made +895% trading NVIDIA and still failed the gate, for taking too much risk to get there. We minted the credential showing exactly that.

A vendor buries the line that fails. We lead with it. A credential engineered to fail itself in public, on its maker's own strategy, is the only kind you can believe when it passes. (Overfitting- and leakage-specific tests — PBO and combinatorial purged cross-validation, both from the same López de Prado framework — are now live too: as of 2026-07-22 all three of his corrections mint as real STARK credentials, re-verifiable at mizan.market/verify. The full framework is now cryptographic proof, a first, and still no strategy of ours has cleared all three bars: the flagship passes PBO and CPCV but its deflated Sharpe reads not significant, and we show that.) The honesty is not a slogan: we publish 9+ of our own retracted results, each with its root cause, because the discipline of catching our own errors is exactly what allocators are paying for. And it is already pulling the right people in — an external quant from an algorithmic-trading firm ran the live engine unprompted.

• Zero-knowledge went production. STARKs are fast and cheap enough to verify a real backtest in milliseconds — impossible a few years ago.
• Systematic capital exploded. More managers, more models, more AI-generated strategies, and a widening trust gap between claim and reality.
• Verifiability became the demand. After a decade of inflated backtests and AI snake-oil, "prove it, don't tell me" is becoming table stakes. The neutral layer that says "this is real" does not exist yet.

Every era of finance installs a verification standard once the market outgrows handshakes. 1909: the first credit rating. 1923: the first audit standard. 1999: GIPS. Systematic finance's standard is being installed now — GIPS verifies the process; MIZAN proves the math.

Trust infrastructure consolidates. One name becomes the standard, and every participant routes through it. MIZAN's moat is three-layered: the ZK proof primitive (hard to build, harder to fake), radical honesty (the self-retraction discipline no competitor will copy), and the neutral position: MIZAN verifies, it does not compete with the managers it certifies.

This is a network-effect business. Each verified manager who gets funded makes the credential more valuable to the next allocator; each allocator who trusts it makes the standard more inevitable to the next manager. Verification layers, once adopted, become the toll on the entire flow they sit on: and MIZAN's flow is verified alpha, which scales with the whole systematic-capital market, not any single fund.

Own the credential, and you own the routing of capital to proven edge.

Where we are, honestly: the engine that mints and verifies the credential is done and live, and the product around it is built and tested: a self-serve manager portal (signup → submission → issued credential → public page), browser-side seal-verification, and a marketplace engine that handles allocations, capital-routing, and fee accounting. What this round buys is distribution and real capital flowing through it: pre-revenue, no customers or live payments yet. The proof is the wedge; the marketplace is the company.

$1.5M seed, on a SAFE. Terms shared directly with committed investors.

The product is built and tested — solo, before a dollar was raised: the engine, the self-serve manager portal, browser-side seal-verification, and the marketplace engine. This is the only external round before revenue, and it is priced to close: distribution, our data feeds, an independent audit, and the first design partners. The hard, irreversible work is behind us. What remains is distribution, and distribution is what this round buys.

Standards get installed once. The founder who owns the credential owns the routing of capital to proven edge — for the life of the category.