Smooth Brains AI — institutional-grade evidence for allocators and third-party verification
Architecture & governance
Architecture, validation methodology, governance, risk management, security model, and fee structure. Does not include proprietary strategy parameters.
Overview for partners & allocators
Pitch deck covering the product, non-custodial model, published performance, and partnership opportunities.
Verify performance yourself
Step-by-step guide to artifact-backed metrics on /performance, hash attestation, and what third parties can independently confirm.
Manifest attestation + robustness bands
How the strategy fingerprint and manifest hash pin published bundles, which files third parties can verify, and how to read Monte Carlo robustness summaries.
Deflated Sharpe summary
Observed Sharpe, trade count, skew/kurtosis, honest variant-count rationale, and PSR/DSR tables for the published fixed FULL window.
Three-layer pipeline summary
Macro and tactical permission layers, automated execution, frozen historical validation, non-custodial model, and how live paper tracking relates to published backtest.
Simulation vs live execution
What the published backtest assumes, how live orders and protective stops behave on Hyperliquid, and where honest gaps remain between simulation and venue fills.
When permission blocks, what happens to open trades
Why existing positions are not force-flattened on every tactical permission flicker, and how protective stops and macro veto still govern tail risk.
Allocator due-diligence grades
Honest weighted composite across edge validation, risk architecture, execution, governance, transparency, capacity, and live track record.
Zero-drift exit-discipline study
Full-history replay isolating the live real-time protective-stop trailing: whether it helps or hurts risk-adjusted performance, and whether it causes live results to drift from the published track record. Concludes by retiring real-time trailing in favour of hourly-close trailing for zero paper-vs-live drift.
Independent AI red-team summary
Executive summary of the July 2026 adversarial hardening audit: scope, findings by severity, capital-at-risk tagging, and remediation status across the execution engine, stop stack, signal pipeline, and published metrics. No exploit paths or secrets.
Why the July 2026 release is better, with evidence
What changed in strategy v7 (state-aware entry gates on the bear-market short satellite), the in-sample and out-of-sample evidence behind the promotion, the ~44 rejected variants, and the honest trade-off in headline return versus risk-adjusted quality.
Independent security review summary
Executive summary of the March 2026 security review: findings, severities, and remediation status. No exploit paths or secrets.
Verified venue round-trips
Public-safe bundle with live metrics, parity status, and artifact hashes once minimum verified round-trip criteria are met.
Live-vs-replay consistency proof
Completion record of the live validation soak: the consecutive clean days during which the live paper-trading mirror matched the rolling research replay, the macro-protection (blocked-trade) day requirement, and the pass result. Published only when the soak genuinely completes.