NC NOETHER CAPITAL
v1.0.0
KM
Noether Capital · Confidential

HIRO Business Plan 2026–2027

Industrial Micro-Alpha Extraction — Engineering Toward 300% Annual Return
Version: 1.0 Date: March 2026 Initial Capital: $1,000,000

All monetary figures denominated in United States Dollars (USD) unless otherwise stated.

1 — Strategic Thesis

1.1 — What Noether Capital Is

Noether Capital is a private, dark quantitative trading operation. It has no external investors, no fund structure, no public visibility. It operates HIRO (Hilbert Intelligent Research & Trading Organism), an AI-native algorithmic trading system designed to industrialise the discovery and harvesting of low-capacity, high-yield, structurally forced market inefficiencies across crypto perpetuals, equity micro futures, commodity micro futures, and Foreign Exchange (FX) micro futures.

HIRO is not a mini hedge fund running scaled-down versions of institutional strategies. It is a micro-alpha extraction organism — a machine that continuously discovers, validates, deploys, and retires many small, fast-cycling, mechanism-bound operators that exploit structural dislocations too small, too messy, or too transient for large firms to pursue. The operator retains 100% of all profits. There are no management fees, no performance fee sharing, no investor reporting, and no compliance overhead from managing other people's money.

The competitive advantage is not prediction accuracy. It is alpha metabolism — the speed at which the system discovers new edges, deploys capital into them, extracts value, detects decay, kills dead operators, and replaces them. At small capital, that metabolism translates directly into extraordinary return on equity because the opportunity set is vast relative to the capital deployed.

1.2 — Why Small Capital Is an Advantage

Large quantitative firms ($1B+ Assets Under Management (AUM)) are constrained by minimum capacity thresholds, internal approval processes, model governance overhead, and the inability to deploy into opportunities that only support small notional. A $10B fund cannot justify pursuing an edge that absorbs $50K before self-impact degrades it. A $1M firm can extract the full value of that edge and move on.

The global market generates thousands of these micro-opportunities every day — funding rate dislocations, liquidation cascade reversals, session boundary distortions, cross-venue lag, queue depletion snapbacks, post-event normalisation, exchange rule-change effects. Each one is small. Each one is brief. Each one is structurally forced. Collectively, they represent an enormous extractable surface for a system built to harvest them at subscale.

1.3 — The 300% Target

The target is $1M initial capital generating $3M–$4M in net extraction over 12 months, representing 300–400% annual return. This is the engineering target that shapes every architectural decision. It is not a guarantee — it is the design point around which capital velocity, operator count, edge density, and kill speed are calibrated.

300%+
Annual Return Target
30–60
Active Operators
5–8×
Daily Capital Turnover
2 — The Extraction Mathematics

2.1 — Working Backward from the Target

300% annual return on $1M = $3M net extraction across approximately 250 trading days = $12,000 average daily net extraction. But daily averages are misleading. Extraction is concentrated in high-density windows — not uniformly distributed. The realistic model is $3K–$5K on quiet days and $20K–$40K on high-density days, with the majority of annual return coming from the highest-density 40–60 trading days.

2.2 — The Capital Velocity Model

The $1M does not sit in positions all day. It cycles through multiple opportunities. If average holding period is 2 hours and 30 operators are active, the same $1M is effectively deployed 4–8 times per day. Effective daily notional = $4M–$8M.

Variable Conservative Target Aggressive
Active operators254060
Avg trades per operator per day235
Total trades per day50120300
Avg holding period4 hours2 hours45 min
Daily capital turnover12×
Effective daily notional$3M$6M$12M
Required avg net edge per trade40 bps20 bps8 bps
Implied daily net extraction$12,000$12,000$12,000
Implied annual return300%300%300%

The Target column is the design point: 40 active operators, 120 trades/day, 2-hour average hold, 6× daily capital turnover, 20 bps average net edge per trade. This is achievable on crypto perpetuals for mechanism-bound, structurally forced opportunities where the edge comes from market plumbing, not prediction.

Critical Sensitivity

8.3 — Regulatory Compliance (S36A)

Noether Capital operates through a Singapore-registered entity. As a proprietary trading firm (not a Digital Token Service Provider), no MAS licence is required, but the system must comply with tax obligations (17% corporate rate) and maintain auditable records.

HIRO implements automated compliance through S36A:

All trading activity is logged to the standard required by Singapore MAS even though the licence threshold is not triggered. This protects the business if regulatory requirements change or if the operation scales to a point where licensing becomes necessary.

The plan lives or dies on two numbers: average net edge per trade (target: 20 bps) and daily capital turnover (target: 6×). If net edge drops to 10 bps, annual return drops to ~150%. If capital turnover drops to 3×, annual return drops to ~150%. Both degrading simultaneously produces ~75%. These two metrics are the primary Key Performance Indicators (KPIs) from Month 3 onward.

2.3 — Extraction Is Window-Based, Not Time-Uniform

HIRO does not seek consistent daily extraction. It seeks maximum extraction during high-density opportunity windows and accepts near-zero activity during low-density periods.

Window Type Frequency Duration Expected Density Mechanism Families
Funding rate reset3× daily (8h cycle)15–45 min around resetHighMP-02
Liquidation cascade2–5× weekly5–30 minVery highFF-02
Session boundary handoffDaily (Asia→Europe→US)30–60 min per handoffMedium-highMP-05
Post-news microstructure normalisationVariable5–60 min after eventHighIT-01
Cross-venue dislocationMultiple dailySeconds to minutesVery high (brief)IL-03
Queue depletion / refillContinuous during volatilitySeconds to minutesMediumLS-01, LS-02
Exchange rule changeMonthly (irregular)Hours to daysVery high (rare)MP-07, MP-08
Contract roll / expiryMonthly/quarterlyHoursMedium-highMP-04

The aggregate of these windows across 20+ symbols provides sufficient opportunity surface. Not every window produces edge. But enough do, frequently enough, to sustain the target extraction rate when the system is properly calibrated.

3 — Architecture Summary

3.1 — The HIRO System

HIRO is specified in the Genesis Framework — 77 sections covering every layer from data ingestion to execution. The system discovers alpha through 20+ parallel discovery engines, validates it through multi-stage testing (statistical, counterfactual, ecological), deploys it as operators with dynamic Capital Fractional Unit (CFU) allocation, executes through alpha-aware routing, and retires operators when edge decays. A full architectural summary is available in the Genesis Framework document.

In Year 1, AI inference for Genesis hypothesis generation and adversarial review is provided via frontier model APIs (Claude, GPT-4o). These are used on the research path, not the execution path — API latency is irrelevant because outputs are operator specifications that undergo hours of validation before deployment. By Year 2, if monthly API spend exceeds the cost of a dedicated GPU server (~€200–400/month), the system transitions to self-hosted open-weights models fine-tuned on 12+ months of HIRO operator history, mechanism taxonomy, validation outcomes, and failure modes. This eliminates external data exposure and produces a model specifically calibrated to HIRO's domain. The decision trigger is purely economic: when self-hosted cost < API cost at comparable quality, switch.

3.1.1 — How HIRO Is Built: The Hilbert AI Software Factory

HIRO is not hand-coded. It is constructed by the Hilbert AI Software Factory — an AI-driven build system that converts the Genesis Framework specification into verified, tested, deployed code.

The Factory operates on Helsinki (Server B) and uses a dual-AI verification pipeline: Claude (defence) proves code matches specification, GPT-4o (prosecution) actively searches for errors. Code must pass BOTH to be deployed. For Rust execution components (L6), a mandatory 7-stage quality pipeline runs before AI verification: cargo check → clippy → test → Miri → audit → bench → fmt.

This approach provides three advantages over traditional development:

The Factory specification is 59 sections. The Code Creator dashboard tracks 97 build packets across 10 construction steps.

3.2 — Infrastructure: Four-Server Optimal Architecture

Server Location Specification Role Cost
A1 (Binance Execution)AWS Osaka (ap-northeast-3)Bare-metal or dedicated — 8+ cores, 64GB+ RAM, NVMe, 1Gbps+Primary Centralised Exchange (CEX) execution. All Binance orders, market data ingestion, Guardian, Risk Envelope, Position Reconciliation, MSE. <5ms to Binance matching engine.~$300/mo
A2 (Bybit + DeFi Execution)AWS Singapore (ap-southeast-1)Bare-metal or dedicated — 8+ cores, 64GB+ RAM, NVMe, 1Gbps+Secondary CEX execution (Bybit, OKX). All DeFi execution — flash loans, oracle liquidations, CEX-DEX arb, Jito Asian bundles. <5ms to Bybit. <100ms to Solana validators.~$300/mo
B (Research)Hetzner Helsinki (AX102)AMD Ryzen 9 7950X, 16 cores, 128GB DDR5, 2TB NVMeAll discovery engines (continuous), Language Engine, Causality Engine, Ecological Simulation, Backtesting, Genesis, Portfolio Optimiser. CPU-heavy, latency-irrelevant.~€160/mo
C (European DeFi Maximal Extractable Value (MEV))Hetzner FrankfurtDedicated — 8+ cores, 64GB RAM, NVMeEuropean DeFi MEV execution. <5ms to Jito Frankfurt relay. Flashbots Protect. When Solana slot leader is a European validator, C fires the bundle instead of A2.~€160/mo

3.2.1 — Why Four Servers

Each server exists because it is the fastest path to a specific extraction venue. A1 in Osaka is <5ms to Binance — the deepest liquidity, most liquidation data, and highest funding volume in crypto. A2 in Singapore is <5ms to Bybit and <100ms to Solana validators — covering the second-largest exchange and all DeFi execution. B in Helsinki provides maximum compute per dollar for CPU-heavy discovery. C in Frankfurt is <5ms to Jito's European relays — when the Solana slot leader is a European validator, C gets the bundle included faster than A2 can from Singapore.

Five servers would produce diminishing returns. Four covers every latency-sensitive destination with <5ms or near-optimal RTT. The marginal gain from a fifth server is single-digit milliseconds on edge cases — not worth the management complexity.

3.2.2 — Cross-Domain Signal Flow

The four-server architecture creates HIRO's primary competitive advantage: cross-domain speed. When A1 (Osaka) detects a liquidation cascade on Binance, it simultaneously executes the CEX reversal trade (<5ms) AND signals A2 (Singapore) via direct tunnel (~70ms Osaka→Singapore) that a DeFi flash loan opportunity may be opening. A2 cross-references against on-chain oracle state (Helius Remote Procedure Call (RPC)) and fires the flash loan via Jito if the oracle hasn't updated yet. Total time from Binance price move to flash loan submission: ~100ms. Pure DeFi bots that only watch the blockchain are still waiting for the oracle to update.

When the Solana slot leader is a European validator, A2 signals C (Frankfurt) instead. C submits the Jito bundle at <5ms to the Frankfurt relay. A2 and C coordinate via direct tunnel (~180ms) to avoid duplicate submissions — the one with the lower-latency path to the current leader fires.

3.2.3 — Inter-Server Communication

Link Round-Trip Time (RTT) Protocol Traffic Type
A1 (Osaka) ↔ A2 (Singapore)~70msWireGuard Virtual Private Network (VPN)Cross-domain signals: cascade detected → fire flash loan. Time-sensitive. Encrypted.
A2 (Singapore) ↔ C (Frankfurt)~180msWireGuard VPNMEV coordination: which server submits the bundle based on current slot leader location. Encrypted.
B (Helsinki) ↔ A1 (Osaka)~250msWireGuard VPNResearch→Execution: operator deployments, parameter updates, CFU changes. Not time-critical.
B (Helsinki) ↔ A2 (Singapore)~250msWireGuard VPNResearch→DeFi: DeFi operator deployments, flash loan strategy parameters. Not time-critical.
B (Helsinki) ↔ C (Frankfurt)~30msWireGuard VPN (or Hetzner vSwitch if same DC)Research→MEV: European DeFi operator updates. Low latency — both in Europe.

3.2.4 — Latency Comparison: Helsinki vs Optimal Four-Server

Destination From Helsinki (old) From Optimal Server Improvement
Binance matching engine (Osaka)~250ms<5ms (A1 Osaka)50×
Bybit matching engine (Singapore)~280ms<5ms (A2 Singapore)56×
OKX (Hong Kong / Singapore)~260ms<30ms (A2 Singapore)
Solana validators (Asia)~280ms<100ms (A2 Singapore)
Jito relay (Frankfurt)~30ms<5ms (C Frankfurt)
Jito relay (Amsterdam)~40ms~10ms (C Frankfurt)
Flash loan: Binance cascade → on-chain liquidation~500ms+~100ms (A1→A2 pipeline)

3.2.5 — Infrastructure Cost

Phase Servers Active RPC / Data Monthly Total
Phase 1 (Month 1–4)A1 Osaka ($300) + B Helsinki (€160)Alpaca $99~$575
Phase 2 (Month 5–6)+ A2 Singapore ($300)+ Helius $200, Alchemy $100, data feeds $350~$1,525
Phase 3 (Month 7+)+ C Frankfurt (€160)Same~$1,700

Full four-server infrastructure at Phase 3: ~$1,700/month or ~$20K/year. On target extraction of $3.5M+/year, that is 0.6% of revenue. Each server pays for itself if it captures just one additional opportunity per month that the previous architecture would have missed.

3.2.6 — Infrastructure Monitoring: The Visual Mesh (S39A)

The 4-server architecture is monitored by the Visual Infrastructure Mesh (S39A), a real-time topological map showing:

If the Osaka-Singapore corridor degrades, the system automatically stalls only the operators dependent on that specific corridor — cross-venue arb pauses but DeFi execution on Singapore continues independently.

Corridor baselines are maintained in configuration and updated when infrastructure changes are detected by the Horizon Scanning module (S40).

3.3 — Instrument Universe

Tier Phase Symbols Rationale
Tier A: Core HarvestMonth 1+BTC, ETH, SOL perpetualsDeepest mechanism density. 24/7 trading. Highest edge density per hour. Concentrate the machine here first.
Tier A: Expanded CoreMonth 3+BNB, XRP, AVAX, DOGE perpetualsSecondary crypto with sufficient liquidity and distinct microstructure characteristics.
Tier B: OpportunisticMonth 5+Micro E-mini S&P 500 (MES), Micro E-mini Nasdaq-100 (MNQ), Micro Gold (MGC), Maximum Capital at Live risk (MCL), M6E micro futuresMonitored continuously. Activated during strong mechanism windows (session opens, macro events, roll periods).
Tier C: Research ReservoirMonth 6+M2K, Micro E-mini Dow (MYM), Micro Silver (SIL), Micro Copper (MHG), M6B, Micro Japanese Yen (MJY), M6A + expansion candidatesWatched for newborn alpha and structural shifts. Not actively traded unless density justifies activation.
4 — Mechanism-First Strategy

4.1 — Operators Are Bound to Structural Mechanisms

HIRO does not trade patterns. It trades structural necessity — situations where market architecture forces price displacement regardless of fundamental value. Every operator in the system is bound to at least one mechanism from the Causality Engine's taxonomy (S24Q). The mechanism tells the operator when the edge should exist, when it should not, and what would cause it to disappear.

The identification of structural mechanisms is not speculative — it is grounded in the UHS Forensics & Analogue Engine (S38). This subsystem reverse-engineers historical episodes of exceptional trading economics, isolates the timeless structural components, and searches current markets for analogous conditions.

The engine maintains three libraries:

This provides HIRO with a disciplined, evidence-based approach to alpha discovery rather than random feature scanning.

4.2 — Primary Mechanism Families for Extraction

Family ID Mechanism Edge Type Expected Contribution
Forced FlowFF-02Margin liquidation cascadePost-cascade reversal or continuation after forced flow exhaustionHigh — most crypto P&L concentration comes from cascade events
Market PlumbingMP-02Funding rate reset distortionPre/post funding rate positioning and convergenceHigh — occurs 3× daily with reliable mechanism
Market PlumbingMP-05Session boundary distortionSession handoff inventory reset, opening range exploitationMedium — daily occurrence, moderate edge per event
Liquidity StructureLS-01/02Liquidity vacuum / passive absorptionPost-sweep snapback, refill pattern exploitationMedium — continuous during volatility
Instrument LinkageIL-03Cross-venue fragmentationShort-horizon dislocation convergence between venuesMedium — frequent but brief
Information TransmissionIT-01Post-news digestion asymmetryOverreaction/underreaction normalisationVariable — depends on news calendar
Market PlumbingMP-07/08Fee / liquidation engine changesNewborn alpha from venue structural changesVery high per event. Frequency increasing as exchanges compete on features. Primary candidates for S24P transient fast-track deployment. Venue Microstructure Change Monitor (S3D.7a) auto-triggers discovery.
Hedging PressureHP-01/04Delta hedging / inventory pressurePredictable directional flow from hedging obligationsMedium — requires options data (Phase 2+)

The top three mechanism families (Forced Flow, Funding Reset, Liquidity Structure) are expected to generate 60–70% of total extraction in Year 1. The remaining families provide diversification and additional extraction during their specific activation windows.

These mechanism families map directly to the UHS Canonical Neighborhood Types defined in S38.9:

Mechanism FamilyS38 NeighborhoodPrimary Discovery Engine
State transmission asymmetryLag-Sync (S38.9.1)S24Q Causality Engine
Forced non-economic flowForced Exit (S38.9.2)S24P Transient Fast-Track
Venue/fee rule changesPlumbing Transition (S38.9.3)S40 Horizon Scanning → S24P
Delayed price incorporationInformation Bottleneck (S38.9.4)S24A Market Behaviour
Small-capacity structural edgeCapital-Fragmented (S38.9.5)S24A, S24P
Market maker inventory imbalanceInventory Stress (S38.9.6)S24Q Causality Engine
Regime transition repricingRegime Transition (S38.9.7)S04 → S24A
5 — Operator Swarm Model

5.1 — Many Small Operators, Not Few Large Ones

The 300% target is not achieved by one or five excellent strategies. It is achieved by running 30–60 operators simultaneously, each capturing a different micro-edge, each with small capital binding, each cycling fast. If each operator averages 2–4 trades per day at $60–$300 net per trade, the aggregate builds.

5.2 — Operator Classes

Class Population Target Avg Lifespan CFU Binding Role
Durable Core10–15 operatorsWeeks to months3–10 Capital Fractional Units (CFUs) eachBaseline extraction. Stable mechanisms. Consistent but modest per-trade edge.
Transient Swarm15–30 operatorsHours to days1–5 CFUs eachCompound Annual Growth Rate (CAGR) acceleration. Short-lived structural windows. High per-trade edge, brief existence.
Symbol-Specific5–15 operatorsDays to weeks1–3 CFUs eachExploit instrument-specific mechanics. BTC funding timing, SOL liquidation dynamics, etc.

The transient swarm is the primary CAGR driver. Durable core provides stability and baseline. Symbol-specific operators capture the highest-conviction, most structurally grounded opportunities on individual instruments.

5.3 — Kill Doctrine

Hard Kill Rule

No operator may hold capital for more than 48 hours without demonstrating positive rolling RAR. If Risk-Adjusted Return (RAR) is negative for 48 consecutive hours, the operator is automatically suspended and its CFUs returned to the reserve pool. The Guardian may grant exceptions for specific operators where the bound mechanism has a documented longer activation cycle, but the default is aggressive kill. Capital trapped in a dead operator is capital not available for a live opportunity. At 300% target, capital velocity is the return driver — every idle dollar degrades CAGR.

6 — Implementation Timeline

5.4 — The Outer Loop: Structural Intelligence (S40)

HIRO does not only search for alpha within the current market structure. The Structural Intelligence & Horizon Scanning module (S40) searches for changes TO the market structure itself — new frictions, new forced flows, new asymmetries.

S40 scans 7 vectors on a dual cadence:

Findings are synthesised by Claude and GPT-4o (adversarial) and converted into Search Directives that feed directly into the discovery engines. This means HIRO can identify and exploit structural changes (e.g., a venue fee change or new DeFi primitive) BEFORE they are visible in price action.

Budget: approximately $30/month for search APIs and AI synthesis.

6.1 — Phase 0: Foundation (Month 1–2)

Target: Zero P&L. System stability. Zero ghost positions. Reconciliation accuracy <1 bps deviation. Attribution engine producing decomposed returns on paper trades. Flash loan contracts tested on testnet. Cross-domain pipeline latency verified <100ms.

6.2 — Phase 1: Controlled Live — CEX + Flash Loans (Month 3–4)

6.3 — Phase 2: Scale to Full Capital + Extended DeFi (Month 5–6)

6.4 — Phase 3: Full Extraction (Month 7–12)

7 — Go/No-Go Gates
Month Milestone Go Gate No-Go Action
2Paper trading stableZero ghost positions. Reconciliation <1 bps. 10+ operators validated. Attribution operational.Fix infrastructure. No live capital until all gates pass.
3First month live ($100K)Avg net edge >15 bps/trade. Max drawdown <5% deployed. Attribution shows signal alpha > execution drag.Reduce to $50K. Diagnose edge vs execution vs data.
4Second month liveCumulative P&L positive. Discovery producing 2+ validated operators/week. Capital velocity >3×.Pause expansion. Focus on pipeline quality.
5Scale to $500KMonthly P&L >$30K on $500K (>70% annualised). Operator count >20. 3+ mechanism families active.Hold at $500K. Investigate concentration risk.
6Scale to $1MMonthly P&L >$50K on $1M. Capital velocity >5×. Attribution clean (residual <15%).Hold at current level. Diagnose velocity or edge issues.
9Full extraction modeRolling 3-month annualised return >150%. Max drawdown <10% NAV. 30+ operators active.If <100%: review mechanism families. If <150%: strong business, below 300% target.
12Year-end assessmentCumulative return >200%.If 150–200%: exceptional outcome. If 100–150%: strong outcome. If <100%: review strategic thesis.
Critical Insight

The go/no-go gates are designed so that underperformance is diagnosed and addressed early, not tolerated. But they also acknowledge that 150% annual return is an exceptional outcome and 100% is still excellent. The 300% target shapes the architecture. The gates ensure the system is producing real value even if the target is not fully reached.

8 — Risk Management

8.1 — Risk Limits

Limit Threshold Action
Per-operator drawdown1.5% of operator's allocated CFUsOperator killed. CFUs returned to reserve.
Per-symbol drawdown3% of total Net Asset Value (NAV)All operators on that symbol paused. Symbol enters HOLD.
Aggregate daily drawdown5% of NAVL3 Global Kill. All positions flattened. Trading suspended until next session with operator sign-off.
Aggregate weekly drawdown8% of NAVSystem enters review mode. All transient operators retired. Only durable core permitted. Full diagnostic review required.
Maximum single position3% of 15-minute average volume (MPS)Pre-order validation rejects oversized orders.
Transient operator CFU ceiling15% of total deployed CFUsNo new transient operators activated until utilisation drops.
Volume-Synchronised Probability of Informed Trading (VPIN) toxicity>0.80 on target symbolNo new orders. Passive orders cancelled. Transients paused.
Aggregate symbol exposure (Anti-Coalescence)Total exposure across ALL active operators on any single symbol must not exceed 0.5% of that symbol's 1-minute median volumeNew orders on that symbol blocked until aggregate exposure drops below threshold. Prevents the swarm from collectively becoming detectable even though each operator individually stays under MPS. This is the quantitative "ghostability" constraint.
Temporal Jitter (Anti-Pattern Detection)When multiple operators fire on the same symbol within the same mechanism window, order submission is staggered across a randomised 200–500ms jitter windowPrevents exchange pattern-matching from linking coordinated orders as swarm behaviour. Each operator's submission time is randomised within the window. No two orders on the same symbol may be submitted within 50ms of each other. Mandatory for all mechanism windows where 3+ operators fire simultaneously.

8.2 — Operational Risk

Risk Mitigation
Ghost positions60-second position reconciliation (S36.4). Any mismatch = immediate alert + order suspension.
Server failureTwo-server architecture. Server A failure: all trading stops safely. Server B failure: trading continues with cached parameters.
Exchange outageMulti-venue routing (S30B). Venue Health Score (S31) auto-pauses on degraded venues.
Data feed failureProvider failover (S3A.9). Stale data detection at 30-second threshold (S36.9). Affected symbols enter HOLD.
Alpha decay faster than replacementContinuous discovery on Server B. Operator birth rate tracked by S24T. Alert if birth rate < death rate for 7 consecutive days.
9 — Key Performance Indicators

9.1 — Primary KPIs (Monitored Daily)

Key Performance Indicator (KPI) Target Alert Threshold Source
Average net edge per trade≥20 bps<15 bps for 5 consecutive daysS35A Attribution
Daily capital turnover≥5×<3× for 3 consecutive daysS28 CFU utilisation
Active operator count30–60<20 for 3 consecutive daysS24T Orchestration
Operator birth rate≥2 per week<1 per week for 2 consecutive weeksS24T Discovery pipeline
Transient operator rotation velocity≥3 births + deaths per day<1 per day for 5 consecutive daysS24P metrics
Signal alpha / total return ratio≥60%<40% (implies returns are from tailwind, not signal)S35A Attribution
Execution drag / signal alpha ratio≤30%>50% (execution consuming too much edge)S35A Attribution
Max drawdown (rolling 30-day)<8% NAV>5% NAV triggers reviewS32 Risk Envelope

These KPIs are monitored through the Genesis Numerical Diagnostics & Integrity Dashboard (S39, GNDID). The GNDID provides real-time visibility across four layers:

  1. Discovery Funnel — is the system finding alpha? (throughput metrics)
  2. Infrastructure Heartbeat — are the servers healthy? (latency, jitter, packet loss)
  3. Execution & Risk Gates — why are trades being rejected? (the “Kill Feed”)
  4. Portfolio & Attribution — how is the money being made? (NAV, P&L, RAR)

Design target: the Research Director can answer “Why aren’t we trading?” within 15 seconds of opening the GNDID, without looking at raw server logs.

Manual parameter tuning is available through the GNDID Fine-Tuning Interface: α (Aggression), β (Friction Buffer), γ (Decay Sensitivity), δ (SNR Threshold), ε (Correlation Ceiling). Guardian limits (S33) cannot be overridden.

9.2 — Extraction Density Dashboard

A real-time panel compositing signals from S24Q (mechanism activations), S24N (language state contradictions), S31.10a (VPIN), S04 (MSE regime), and Venue Health Score into a single ranked view showing per-symbol opportunity density at any moment. The top 5–10 symbols are where capital should be concentrated. This is the primary operational interface for managing the extraction engine.

10 — Financial Projections

10.1 — Scenario Analysis

Scenario Avg Net Edge Capital Velocity Year 1 Net Return Year-End NAV
Bear Case10 bps~75%$1.75M
Base Case15 bps~190%$2.9M
Target Case20 bps~300%$4.0M
Bull Case25 bps~500%$6.0M

10.2 — Profit Sweep & Capital Tiering

HIRO's Tier 1 extraction engine operates on a fixed $1M capital base. As profits accumulate, they are swept to a Tier 2 capital reserve on a defined schedule. This serves three purposes: it maintains the ghost footprint (keeping Tier 1 capital small), it protects realised profits from Tier 1 drawdown events, and it builds a capital reserve that can fund Tier 2 strategies, re-inject into Tier 1 during extreme opportunity windows, or be withdrawn as income.

10.2.1 — Sweep Rules

Rule Specification
Sweep frequencyWeekly (every Friday UTC close). Profits above the $1M Tier 1 base are swept to Tier 2.
Minimum sweep$5,000. If accumulated weekly profit is below $5K, it remains in Tier 1 until the next sweep.
Tier 1 replenishmentIf Tier 1 NAV drops below $900K due to drawdown, Tier 2 injects capital back to restore $1M. This is automatic up to $100K per injection. Larger replenishments require Guardian approval.
Tier 1 maximum$1.2M. If intra-week profits push Tier 1 above $1.2M, an immediate mid-week sweep is triggered. This prevents Tier 1 from growing beyond the ghost footprint threshold.
Tier 2 deploymentTier 2 capital is held in stable assets (USDT/USDC on exchange, or stablecoin yield if available above risk-free). Tier 2 is not actively traded in Year 1 — it is a reserve and profit accumulation vehicle.

10.2.2 — Capital Growth by Scenario (Year 1)

Month Bear (75% Annual Return (AR)) Base (190% AR) Target (300% AR) Bull (500% AR)
1–2Phase 0: Paper trading. No P&L. Infrastructure build.
3$1,010K$1,025K$1,040K$1,060K
4$1,020K$1,055K$1,085K$1,130K
5$1,040K$1,100K$1,170K$1,280K
6$1,070K$1,170K$1,310K$1,520K
7$1,110K$1,260K$1,500K$1,830K
8$1,150K$1,360K$1,720K$2,200K
9$1,200K$1,480K$1,980K$2,650K
10$1,260K$1,620K$2,280K$3,200K
11$1,330K$1,780K$2,630K$3,870K
12$1,400K$1,960K$3,030K$4,700K

Note: total capital = Tier 1 ($1M fixed) + Tier 2 (accumulated sweeps). Growth shown above is total capital across both tiers. The Tier 1 extraction engine always operates on ~$1M. Profits accumulate in Tier 2.

Month 3–4 shows slower growth because only $100K is deployed (Phase 1 Controlled Live). Full $1M deployment begins Month 5–6. The acceleration from Month 7 onward reflects full capital velocity on all mechanisms.

10.2.3 — Tier 2 Capital Deployment

Noether Capital is a private, dark quant operation. There are no external investors, no fund management obligations, no regulatory overhead from managing other people's money. All profits are retained. The entire organisational drag that slows institutional firms — investor reporting, compliance committees, marketing, lock-up structures — does not exist. This is a structural advantage that compounds over time.

As Tier 2 capital accumulates from weekly profit sweeps, it is deployed through a self-compounding architecture:

Tier 2 Balance Deployment
$0–$250KPure reserve. Held in stablecoins (USDT/USDC) on exchange or in yield-bearing stablecoin protocols if yield exceeds 5% Annual Percentage Rate (APR) with acceptable counterparty risk. Primary function: Tier 1 drawdown replenishment buffer. No active trading.
$250K–$750KReserve ($250K maintained) + second Tier 1 extraction engine ($500K–$1M). The second engine runs on the same infrastructure (Server B handles discovery for both engines) but operates on a different symbol subset — for example, Engine 1 on BTC/ETH/SOL, Engine 2 on secondary crypto + equity micro futures. Each engine maintains its own ghost footprint. Combined extraction doubles without increasing per-symbol visibility.
$750K–$2MTwo Tier 1 engines ($1M each) + Tier 2 active strategies on remaining capital. Tier 2 strategies are longer-horizon, higher-capacity: basis arbitrage (perp vs spot), cross-venue carry, funding rate harvesting at scale, tail-risk hedging. Target: 30–80% AR on Tier 2 capital. Lower velocity, higher capacity, different mechanism families.
$2M–$5MThree Tier 1 engines ($1M each, different instrument universes) + Tier 2 active ($2M+). Add third server for third extraction engine. Total extraction surface spans crypto, equity, commodity, and FX across three independent ghost-footprint engines. Personal income withdrawal enabled from Tier 2 without reducing any Tier 1 engine capital.
$5M+Maximum self-funded scale. Four to five Tier 1 engines across diversified instrument classes and geographies. Tier 2 at $2M–$3M in active strategies. Remaining capital in long-term wealth preservation (stablecoin yield, treasury instruments, physical assets). Total annual extraction potential: $5M–$15M depending on market conditions. No employees. No investors. No overhead beyond infrastructure and data.

The dark fund model compounds faster than an institutional model because there is no management fee drag, no performance fee sharing, no investor redemption risk, no compliance overhead, and no organisational complexity. Every dollar of profit is available for redeployment. The trade-off is that growth is limited to retained earnings — but at 300% AR on Tier 1, retained earnings grow fast.

10.3 — Complete Cost Structure

Cost Category Detail Monthly Annual
Infrastructure
Server A1 (Binance Execution)AWS Osaka bare-metal — co-located with Binance matching engine~$300$3,600
Server A2 (Bybit + DeFi)AWS Singapore bare-metal — co-located with Bybit, DeFi execution~$300$3,600
Server B (Research)Hetzner Helsinki AX102 — Ryzen 9 7950X, 128GB DDR5, 2TB NVMe€160 (~$175)$2,100
Server C (EU DeFi MEV)Hetzner Frankfurt — dedicated, near Jito relay€160 (~$175)$2,100
Inter-server VPNWireGuard mesh: A1↔A2↔B↔CIncluded$0
Domain + Secure Sockets Layer (SSL)Dashboard hosting, monitoring endpoint~$10$120
Infrastructure subtotal$960$11,520
Data Feeds
Alpaca MarketsUnlimited plan — US equity real-time, Securities Information Processor (SIP) feed$99$1,188
LunarCrushCrypto social metrics — Phase 1~$100$1,200
SantimentCrypto social + on-chain — Phase 1~$50$600
Benzinga ProNews feed API — Phase 1~$80$960
Trading EconomicsMacro calendar API — Phase 1~$30$360
CryptoPanicCrypto news aggregator — Phase 1Free tier$0
Wikipedia PageviewsWikimedia REST API — Phase 1Free$0
FREDFederal Reserve macro data — Phase 1Free$0
Binance / BybitMarket data + liquidation streams — WebSocketFree$0
Polygon.io (Phase 2)Equities L2 — when equity micros go active~$80$960
Helius Solana RPCTrader Node — premium Solana RPC with Jito bundle support + Yellowstone gRPC streaming~$200$2,400
Ethereum RPC (Phase 2)Alchemy or Infura — premium Ethereum Virtual Machine (EVM) RPC with Flashbots Protect~$100$1,200
Data feeds subtotal$739$8,868
AI & Software
Claude APIGenesis, adversarial review, hypothesis generation~$150$1,800
GPT-4o APIAdversarial auditor (S11B dual review)~$50$600
Self-hosted AI (Year 2+)Replaces API costs when monthly API spend exceeds GPU server cost. Hetzner GPU server ~€200–400/month running fine-tuned open-weights model.$0 (Year 1)$0 (Year 1)
Software licensesMisc tools, monitoring, backup~$30$360
AI & software subtotal$230$2,760
Factory & Diagnostics (Build Phase)
Software Factory — AI verificationClaude (defence) + GPT-4o (prosecution)0–100/day00–,000
GNDID — AI diagnosticsStall analysis + health reports (S39).40/day76
Horizon Scanning — SIHSSearch APIs + AI synthesis (S40)~060
Rust Quality Pipelinecargo audit, Miri, benchmarks — runs locally/usr/bin/bash/usr/bin/bash
Factory & diagnostics subtotal~20~,440
Trading Costs (Variable)
Exchange fees (standard tier)Maker 0.02% / Taker 0.04% on Binance. Blended ~0.03% at 60/40 maker/taker mix.VariableSee below
Exchange fees (VIP tier)After qualifying: Maker 0.016% / Taker 0.036%. Blended ~0.024%.VariableSee below
SlippageEstimated 1–3 bps per trade on crypto perpetuals at HIRO's position sizesVariableSee below
Funding rate (net)Net funding received minus paid. Can be positive (income) or negative (cost). Historically net positive for short-bias strategies.VariableSee below

10.4 — Full Profit & Loss by Scenario (Year 1)

Assumes: Month 1–2 no trading (Phase 0). Month 3–4 at $100K deployed. Month 5–6 scaling to $1M. Month 7–12 full extraction. Exchange fees at VIP tier from Month 4 onward (volume qualification). All figures in USD.

Line Item Bear (75%) Base (190%) Target (300%) Bull (500%)
Revenue
Gross trading P&L$820,000$2,080,000$3,300,000$5,500,000
Maker rebates earned$8,000$18,000$28,000$45,000
Net funding income$5,000$12,000$20,000$30,000
Total gross revenue$833,000$2,110,000$3,348,000$5,575,000
Trading Costs (Variable)
Exchange fees (taker)($28,000)($65,000)($100,000)($165,000)
Exchange fees (maker, net of rebate)($4,000)($9,000)($14,000)($22,000)
Slippage cost($35,000)($80,000)($125,000)($200,000)
Funding cost (periods of net payment)($12,000)($25,000)($38,000)($55,000)
Total trading costs($79,000)($179,000)($277,000)($442,000)
Fixed Costs
Infrastructure($4,320)($4,320)($4,320)($4,320)
Data feeds($5,268)($5,268)($5,268)($5,268)
AI & software($2,760)($2,760)($2,760)($2,760)
Total fixed costs($23,148)($23,148)($23,148)($23,148)
Summary
Net profit (Year 1)$730,852$1,907,852$3,047,852$5,109,852
Net return on $1M~73%~191%~305%~511%
Trading cost as % of gross9.5%8.5%8.3%7.9%
Fixed cost as % of gross2.8%1.1%0.7%0.4%
Tier 1 year-end NAV$1,000,000$1,000,000$1,000,000$1,000,000
Tier 2 year-end balance$730,852$1,907,852$3,047,852$5,109,852
Total capital year-end$1,730,852$2,907,852$4,047,852$6,109,852
Key Insight

Trading costs (exchange fees + slippage + funding) represent 8–10% of gross revenue across all scenarios. This is the dominant cost. Fixed costs (infrastructure + data + AI) represent only 0.2–1.5% of gross. The business has extreme operating leverage — nearly all incremental gross revenue drops to the bottom line. The most impactful cost optimisation is exchange fee tier negotiation and maker/taker mix improvement, not infrastructure savings.

10.5 — Multi-Year Self-Compounding Projection

Noether Capital operates as a private dark fund. No external capital. No investors. 100% profit retention. Growth is funded entirely from retained extraction. The projection below assumes Tier 1 engines operate at the Base Case (190% AR) after Year 1 — a conservative assumption that accounts for mechanism crowding, market structure evolution, and natural edge decay over time.

End of Year Tier 1 Engines Tier 1 Capital Tier 2 Capital Total Capital Annual Extraction Cumulative Extraction
Year 11$1.0M$2.0M$3.0M$2.0M$2.0M
Year 22$2.0M$4.8M$6.8M$3.8M$5.8M
Year 33$3.0M$9.5M$12.5M$5.7M$11.5M
Year 44$4.0M$15.1M$19.1M$7.6M + Tier 2 income$19.1M+
Year 54–5$4–5M$20M+$25M+$8–10M+ combined$27M+

Assumptions: Year 1 at Base Case (190% AR on $1M = ~$2M extraction, conservative below 300% target). Years 2+ at 190% AR per Tier 1 engine. New Tier 1 engine launched when Tier 2 accumulates sufficient capital ($750K+ surplus above reserve). Tier 2 active strategies begin Year 2 at 50% AR on deployed Tier 2 capital. Personal income withdrawals of $200K/year from Year 2 onward (deducted from Tier 2, not Tier 1). By Year 2, AI inference transitions from API to self-hosted open-weights models fine-tuned on HIRO operator history — eliminating external data exposure and reducing marginal AI cost to near zero.

Infrastructure scales with engine count: each additional Tier 1 engine requires one additional AX102 server (~€160/month) connected via vSwitch. By Year 2, if API spend exceeds ~$1,000/month, a dedicated GPU server (Hetzner GPU line, ~€200–400/month) replaces API calls with self-hosted fine-tuned inference, eliminating external data exposure and reducing marginal AI cost to near zero. Total infrastructure cost at full 4–5 engine scale: ~€1,200–1,500/month ($15K–$18K/year) — still negligible relative to extraction.

Scaling Constraint

Each Tier 1 extraction engine requires its own ghost footprint — operating on a distinct symbol subset to avoid cross-engine coalescence. Engine 1 might run BTC/ETH/SOL. Engine 2 might run secondary crypto + equity micros. Engine 3 might run commodity + FX micros. By Year 3–4, the symbol universe is fully covered across multiple engines, and further Tier 1 scaling requires expanding into new instrument classes or new venues. Beyond 4–5 Tier 1 engines, additional capital is better deployed in higher-capacity Tier 2 strategies.

10.5.1 — Personal Income & Wealth Building

Year Estimated Personal Income Source Notes
Year 1$0All profits retained for compounding. Living expenses funded from existing savings.
Year 2$200K–$400KTier 2 withdrawalFirst income year. Withdrawn from Tier 2 surplus above engine funding requirements.
Year 3$400K–$800KTier 2 withdrawal + Tier 2 yieldTier 2 balance supports both engine funding and substantial personal income.
Year 4+$500K–$1M+Tier 2 income + capital gainsSystem self-sustaining. Personal income no longer constrains capital growth.

By Year 3, the system is generating sufficient extraction that personal income withdrawals do not constrain Tier 1 or Tier 2 capital growth. The dark fund model means no management fee drag, no performance fee sharing, no fund administration costs — the operator keeps everything.

10.5.2 — Why Dark

Noether Capital deliberately operates without external capital, external investors, or public visibility. This is not a temporary constraint — it is the permanent strategic posture. The reasons are structural:

11 — What Makes This Achievable

The 300% target is achievable — not guaranteed, but structurally plausible — because of five properties that compound together:

First, the opportunity surface is real. Crypto perpetual markets generate thousands of structurally forced micro-dislocations daily. Funding resets, liquidation cascades, cross-venue lag, queue depletion events — these are not hypothetical. They are observable, measurable, and mechanically explained. The question is not whether the opportunities exist but whether HIRO can capture them efficiently.

Second, the capital is small relative to the opportunity. At $1M, HIRO's per-trade size is far below the detection threshold of institutional participants and far below the capacity ceiling of most micro-alpha opportunities. The system can enter and exit without moving the market, which preserves the full theoretical edge.

Third, the architecture is purpose-built for this extraction model. HIRO is not a general-purpose trading system adapted for micro-alpha. It is designed from the ground up for high operator count, fast cycling, mechanism binding, continuous discovery, aggressive kill, and dynamic capital routing. Every section of the Genesis Framework supports this operating model.

Fourth, the feedback loops are tight. P&L attribution (S35A) decomposes returns daily. Meta-productivity predictions (S24R) anticipate regime changes. Crowding detection (S24P, S24R, S28.5a) routes capital away from degrading edges. Negative primitive feedback (S33A) prevents repeated failures. The system learns from its own operation faster than most manual research processes.

Fifth, the cost structure is minimal. $23K annual fixed costs on a $1M capital base is 2.3%. Even at the bear case scenario (73% return), the firm is generating $731K on $23K of fixed costs. The four-server architecture — Osaka, Singapore, Helsinki, Frankfurt — puts HIRO within <5ms of every major extraction venue on the planet. Each server pays for itself if it captures just one additional opportunity per month that the previous architecture would have missed.

The 300% target is not a prediction. It is the design point that shapes every architectural decision — operator count, capital velocity, kill speed, discovery cadence, mechanism focus. If the system reaches 200%, it has built an exceptional business. If it reaches 150%, it has built a strong business. If it reaches 100%, it has built a good business. The target ensures the machine is built for maximum extraction. The gates ensure it adapts to whatever reality delivers.

12 — Phase 2 Expansion Opportunities

Phase 1 operates on CEX perpetuals and micro futures PLUS flash loan extraction (oracle lag liquidations via the A1→A2 cross-domain pipeline). Flash loans are Phase 1 because they require zero capital at risk and the infrastructure — CEX price feeds, Solana RPC, flash loan contracts — is built during Phase 0. The remaining DeFi opportunities (CEX-DEX arb, DeFi perp funding, MEV bundles) and additional CEX opportunities (new listings, basis, options vol, stablecoin depeg) activate in Phase 2 (Month 5+) as the DeFi execution adapter matures.

12.1 — DeFi Execution Opportunities

Oracle Lag Exploitation

CEX-DEX Price Arbitrage

DeFi Perp Funding Arbitrage

MEV Bundle Extraction (Solana)

Flash Loan Amplified Extraction

Flash Loan Use Cases for HIRO

All DeFi execution is handled by the Alpenglow DeFi Execution Module (S31C), deployed on Server A2 (Singapore). Alpenglow supports: flash loan routing across multiple providers, AMM swap execution with MEV-aware routing (Jito on Solana), oracle-dependent execution with real-time oracle lag monitoring, gas budget management (flash loan operators consume gas budget, not CFUs), and cross-chain execution across Solana and EVM chains. Oracle lag thresholds: Tier-1 oracles (Chainlink, Pyth) at 300ms, Tier-2 at 500ms — monitored by GNDID Layer 2.

Flash Loan Risk & De-Risk

12.2 — Additional CEX Opportunities

New Listing Exploitation

Cross-Exchange Basis Trade at Scale

Options-Implied Volatility Exploitation

Stablecoin Depeg Exploitation

12.3 — Phase 2 Opportunity Summary

Opportunity Category Conservative Annual Moderate Annual Infrastructure Needed
Oracle lag exploitationDeFi$200K$500KSolana/EVM RPC, liquidation contracts
CEX-DEX arbitrageDeFi$100K$400KDEX aggregator integration
DeFi perp funding arbDeFi$100K$300KDeFi perp protocol APIs
MEV bundle extractionDeFi$100K$300KJito Block Engine API
New listing exploitationCEX$50K$200KExchange announcement monitor
Cross-exchange basis at scaleCEX$100K$300KMulti-venue capital deployment
Options-implied vol exploitationCEX$50K$200KDeribit data feed
Stablecoin depeg exploitationCEX/DeFi$50K$150KStablecoin price monitoring
Flash loan amplified extractionDeFi (PHASE 1)$500K$2.0MAave/Balancer flash loan contracts, tx simulation — deployed Month 1–2, live Month 3
Phase 2 total (additional DeFi + CEX, excl. flash loans)$750K$2.35M
Phase 1 (CEX + flash loans) extraction surface$2.7M$7.0M
Phase 1 + Phase 2 combined surface$3.5M$9.4M

Flash loan extraction is Phase 1 — not deferred. It requires zero capital at risk, uses infrastructure already built for CEX trading (price feeds, server architecture), and adds $500K–$2M in annual extraction potential from Month 3. Combined with CEX extraction ($2.2M–$5.0M), Phase 1 alone covers a $2.7M–$7.0M opportunity surface. The $3M target is achievable from Phase 1 without any Phase 2 expansion. Phase 2 DeFi and additional CEX opportunities add $750K–$2.35M more, pushing the total combined surface to $3.5M–$9.4M.