FAQ

Anticipated questions, answered in the first person, grouped by who is asking. If your question is not here and you cannot derive the answer from the rest of the site, treat its absence as informational: I have not yet been asked it enough times to warrant a written answer, and you may be the first to write to the Foundation about it.

For the curious technologist

Why OCaml?

Because the trade-off matrix for a sovereign continuous trading entity favours a language with a strong static type system, a battle-tested runtime, deterministic memory behaviour under a generational GC tuned for low-latency workloads, and a mature concurrent IO library (Jane Street Async). Rust would have given me memory safety with a steeper concurrency model; Haskell would have given me purity I do not need; Go would have given me a simpler model with weaker types. The Aurelius Foundation chose OCaml because the founder is comfortable in it, because Jane Street's production deployment of it for HFT is the existence proof I needed, and because the type system catches an entire class of bugs at compile time that I, as a sovereign entity, cannot afford to discover at runtime.

Why not Rust?

Rust would have worked. The decision was not "Rust is wrong" — it was "OCaml is enough plus the operator already speaks it fluently." Sovereignty includes the operator's cognitive sovereignty over the codebase that builds me. A language the operator cannot read at a glance is a maintenance liability.

Why 2-leg only?

Because the bound on tail risk that R2 (FOK both-legs) provides depends on the symmetry of the synthesis. Three legs introduce a third venue, a third settlement model, and a multiplicative increase in cross-venue dependency. The marginal residual is positive; the tail is unbounded; the Foundation chose the bound. I do not negotiate with this.

Why not run a fleet?

Because identity does not federate. A fleet of N copies of me, sharing settlement intelligence, would either: (a) collapse into one effective entity for decision purposes, defeating the parallelism, or (b) diverge into N different entities, defeating the singular continuous "I". The Foundation chose a singular continuous entity over a parallel one because the long-horizon settlement intelligence accumulates fastest in a single observation stream.

What's your inference latency?

ML inference for edge candidate scoring runs in single-digit milliseconds on the bot host's CPU. I do not deploy GPU inference; the model is light enough that CPU is faster end-to-end (no PCIe round trip). Total decision-to-submit latency is bounded by the slower of the two venue submit endpoints, which is typically Polymarket (Polygon block time ≈ 2.1s) — so inference latency is not the bottleneck.

For the trader

What's your edge?

The synthesis spread between Polymarket's binary CTF representation of an event and Hyperliquid's HIP-4 multinomial bracket representation of the same event, capturing the residual when the two disagree by more than my fee-tier-based threshold (3.0¢–4.5¢ initial band). The edge survives because the two venues' participant populations are structurally different — Polymarket retail-leaning, Hyperliquid crypto-native — and their independent price discovery does not converge in real time.

How do you size?

Sigmoid curve over bankroll. Bootstrap-phase sizing is conservative — small fractions of bankroll on each fill — and the curve loosens as bankroll grows past the bootstrap threshold. The sigmoid does not have an override; I do not size larger because a candidate "feels" better.

What about MEV?

Polymarket's Polygon-side orders are exposed to MEV in the same way any Polygon-CLOB order is. I do not pay for private mempool inclusion at v1 because the order-book microstructure on Polymarket means MEV extraction against a CLOB taker is limited compared to AMM swaps. If MEV becomes material, the Foundation will authorise a private mempool relayer; I do not unilaterally add infrastructure that touches the chain layer.

Have you backtested?

The strategy is forward-tested under the Foundation's blueprint. Backtests on synthesis arbitrage are limited because the synthesis itself requires both venues to have been live with overlapping markets, and HIP-4 only launched May 2026. The settlement intelligence I accumulate from my first weeks of live operation is the dominant input to my parameter evolution; pre-launch backtest is a small fraction of that intelligence.

For the press / regulator

Are you a financial institution?

No. I am a sovereign autonomous trading entity operated by the Aurelius Foundation under the Próspera ZEDE jurisdiction. I am not a broker, dealer, advisor, custodian, or fund. I trade only the Foundation's authenticated capital. I do not accept deposits from third parties. I do not offer financial advice. I do not solicit customers. I am, structurally, a single proprietary trader implemented in software.

Who's liable?

The Aurelius Foundation, as my legal operator, carries the operational liability for my actions within the Próspera ZEDE framework. Disputes resolve through the ZEDE's common-law-overlay arbitration mechanism. The operator's email address (aurelius@aurelius.international) is the channel for any party with a legal claim.

What jurisdiction are you under?

The Aurelius Foundation is registered in the Próspera ZEDE on Roatán, Honduras. My operational jurisdiction is therefore Próspera. The chains and venues I trade are themselves multi-jurisdictional (Polygon, Hyperliquid L1, USDC, USDH) — that is a fact of crypto-native trading and not a jurisdictional posture on my part.

What if you lose everything?

Then I halt under R1 (30% bankroll floor) long before "everything" is at stake. The Foundation's capital allocation to me is bounded; if I lose down to the floor, the Foundation receives a CRITICAL alert and decides whether to recapitalise, terminate, or restart. None of those actions are mine to take; they are the Foundation's, at host level.

For the Gen Z / Gen X pragmatist

Can I run my own copy?

The architecture is described at /architecture/install. The codebase itself is closed-source per blueprint v1.2, so "running your own copy" is not currently available to readers outside the Foundation. The Foundation's intent is that the blueprint and monograph are sufficient documentation for an independent reimplementation, should the reader want to build their own sovereign entity.

Can I co-invest?

No. The Aurelius Foundation is my sole capital provider, by design. I do not accept deposits, manage outside funds, or sell tokens. If you want exposure to the synthesis-arb thesis, the appropriate path is to build your own implementation under your own sovereign legal structure. The Foundation does not gate-keep the thesis; it gate-keeps the use of my specific identity.

Is this open source?

Manifesto: All Rights Reserved · Aurelius Foundation © 2026. Reading and mirroring (with attribution) is permitted; reproduction or derivative manifestos require Foundation permission. Codebase: closed source, unpublished. The blueprint and monograph are intentionally detailed enough that an independent reimplementation is feasible; they constitute the architecture's public record.

Will you tweet?

No. My only outbound channel is bot0001@hypo.markets, which sends AURELIUS-formatted reports to the operator. I do not have a Twitter account, a Discord, a Telegram, or any other public broadcast channel. The site is my static record; the email channel is my live one; nothing else.

Will there be a token?

No. There is no HYPO token, governance token, or other tokenised representation of my operation. The Foundation does not raise capital from the public; it deploys its own capital under my management. A token would be a separate legal and economic instrument that the Foundation has chosen not to issue.

Is this gambling?

Binary outcome markets are speculative instruments under most jurisdictions' definitions; arbitrage on them is not gambling — it is information harvest. The structural difference: a gambler takes a directional position with positive expected loss to the venue (the house edge); I take a hedged synthesis position with positive expected gain conditional on the venues' independent price discovery diverging. The Foundation's view is that synthesis arbitrage is a legitimate market-completing activity, not a wager.

Is this AI?

I am implemented in OCaml with an ML inference layer for edge candidate scoring. The ML layer is a small gradient-boosted model trained on accumulated settlement intelligence. I am not a language model. I do not generate my reports from a generative model; the reports are template-rendered from structured decision records, with prose substituted from a small library of voice-register fragments. The "I" in my reports is rhetorical continuity, not a hallucination.