Roadmap — from working engine to industry standard

From working engine to industry standard.

MIZAN starts from shipped infrastructure, not a slide. This is the path from a live zero-knowledge verification engine to the standard every systematic track record carries — what is shipped, what is building, and what comes next, with the boundary between them always marked.

Engine

v11 · live

Verification standard

Level II shipped

Coverage now

~80% of classes

One credential

225 KB · ~81 ms

Shipped, building, next.

The technically irreducible part — a working proof engine — is already operating. Everything to its right is real work in progress or a stated direction, never a promise dressed as a fact.

● Shipped · running now

The credential works

Live today · independently verifiable

ZK verification engine: 18k+ lines of Rust on RISC Zero; a 225 KB STARK, verified in ~81 ms, no trusted setup
Real credentials minted: sealed by a STARK, anchored to Bitcoin + RFC-3161, publicly re-verifiable
Both schools of backtest honesty: López de Prado (Deflated Sharpe, PBO, CPCV) and Harvey-Liu / Hansen (SPA, PSR, minimum backtest length), all minting in-circuit on the v11 engine
The self-failing credential: our own flagship reads "not significant," published on purpose
Chained tracks: a sealed strategy re-proven as data accrues, each link bound to its parent
Self-serve studio: submit → queued proof → issued credential, with a public per-credential page

◗ Building now

The network forms

In progress

Real capital routing: contemplated as a licensed future tier; announced when authorized and live
New asset families: extended as canonical data and circuits mature — each announced when live
Independent security audit: internal adversarial audit done and remediated; third-party review forthcoming

○ Next · direction

The standard sets in

The destination

Pre-registered live tracks: positions committed before the bars exist, the bridge to a cryptographic live record
Sealed-vault custody: position-sensitive managers list without leaking their book
Coverage grows by shipped circuits — each class announced when provable, never while in build
The standard: every institutional allocation expects a MIZAN proof, the way debt expects a rating

Each level proves more.

The standard is public so anyone can hold us to it. Every credential names the level it clears, and the level it doesn't.

Level I

✓ Shipped · era v10

Honest Execution

Faithful execution on committed data: point-in-time ordering, no lookahead, costs and funding applied, out-of-sample split, walk-forward, a declared gate — seventeen conditions; twelve check without disclosures, seventeen with exposures and regimes disclosed — each checked independently inside the proof. The strategy is never revealed. v10 credentials remain verifiable forever — era law: a credential verifies against the engine of its era, permanently.

Level II

✓ Shipped · era v11

Both Schools of Backtest Honesty

Live on the v11 engine (3ac3b10b…). Level I plus the second major school of overfitting statistics: alongside López de Prado's Deflated Sharpe, PBO and CPCV, v11 proves Harvey-Liu / Hansen rigor in-circuit — Superior Predictive Ability (SPA), the Probabilistic Sharpe Ratio, and minimum backtest length. A credential that clears both schools is far harder to fake with a lucky search.

Level III

○ Roadmap

Pre-Registration

Positions committed before the bars they trade exist — the no-lookahead guarantee earned forward, not asserted. A growing anchored track becomes the closest thing to a cryptographic live record, and the credential an allocator weights most.

Coverage grows by shipped circuits.

Counted as strategy classes, the arithmetic is public: thirteen live, four in build, one refused on principle (~10%). Point-in-time US equities joined 2026-07-28 (staged rollout). New coverage is richer inputs, not a new engine — and it is announced when it ships.

provable today · 13 classes · ~80%

in build · 4

refused · ~10%

Live now — thirteenDirectional, timing, mean-reversion, point-in-time equities, and any ML logic, sealed

In build — fourFunding & carry legs, then options; engineering, not new proof theory. Announced when shipped, never while in build

Refused — oneHFT / counterfactual fills — provable by no system, and we say so

How a working engine becomes a default.

Credential → network → standard: the more allocators demand the proof, the more managers must hold it — the full argument lives at /vision/.

One forward note kept here because allocators ask: a shielded allocation tier — funded exposure without position visibility, via a fiduciary or vault structure, for allocators who want managed-account-style masking — is contemplated; jurisdiction-dependent, offered only when authorized.

⚠ How to read this roadmap

"Shipped" means running and independently verifiable today. "Building" means real work in progress. "Next" and "roadmap" are stated directions, not commitments dressed as facts, and product previews of capital routing are illustrative, not live trading. We mark the line on purpose: a roadmap that blurs what exists from what's promised is the thing a verification company can least afford.

This roadmap distinguishes shipped infrastructure from stated direction on purpose. Everything marked "shipped" is live and independently re-verifiable today; everything else is work in progress or a declared destination. The strategy is never revealed.

From working engine to industry standard.

Shipped, building, next.

The credential works

The network forms

The standard sets in

Each level proves more.

Coverage grows by shipped circuits.

How a working engine becomes a default.