Public transparency page

An AI fleet manages this portfolio. Every number below is live.

Luxverum Capital is an autonomous, AI-managed equity portfolio in a monitored paper forward-test. Its purpose: prove the engine, then direct profits to a non-profit feeding children in Venezuela. Nothing here is hidden — performance, risk posture, idle-cash yield, fees, and the machine's own audit trail are published in real time.

Download monthly factsheet (PDF) Bilingual EN/ES · auto-generated monthly · fund vs SPY, QQQ & T-bills

Total return since inception

—

Alpha vs S&P 500

—

Downside protection

—

Market regime

—

Growth of capital — fund vs S&P 500

Cumulative return since forward-test baseline · indexed to 0%

Luxverum S&P 500 (SPY)

Alpha vs SPY

Cumulative out/under-performance, tracked daily

Cash never sleeps

Idle cash is swept into a T-bill reserve daily

—

— Deployed + working cash

Un-deployed capital is parked in a 1–3 month T-bill cash-equivalent ETF, earning money-market yield until the trader deploys it. Buys redeem from the reserve automatically — yield never blocks a trade.

Fees — minimal by design

One small all-in fee. No performance cut, no lock-ups, nothing hidden.

Management fee

1%/year

One aggressive all-in fee: frontier AI models, infrastructure, legal & compliance, and operations. Accrued daily on assets under management.

Performance fee

We take no cut of your gains. Returns shown are what the investor keeps, minus the single management fee.

Entry, exit & lock-up

None

No subscription fees, no redemption fees, no minimum holding period.

For context: a typical hedge fund charges 2% management + 20% of profits. Luxverum charges 1% + 0 — half the standard fee and none of your gains. On $10,000 that is $100 per year — about $8.33 per month.

Planned fee schedule for when the fund opens to outside capital. No client funds are accepted today (paper forward-test). Final terms will be set out in the offering documents; in case of any discrepancy, the offering documents control.

How the machine works

Five AI roles, hard-coded risk limits the AI cannot override

The approach

Loading…

Risk controls (enforced in code)

Loading…

How we run this fund

Operating prospectus — stability first, then returns, transparency always

Losses are prevented in code, not by judgment. Every safeguard below is enforced by deterministic software the AI cannot override — the machine can be wrong about a stock, but it cannot skip a stop, breach a cap, or hide a bad day.

Governance — separation of duties

Trader (Sonnet 5, hourly during market hours) finds setups and executes — within limits it cannot change.
Validator (Opus, twice daily) independently audits every decision.
Protector (hourly, read-only) verifies downside protection on every position; it can alert, never trade.
Fleet admin (daily) reviews health, errors and buy-funnel discipline.
Analyst (weekly) measures expectancy and calibration — the go-live gate for real capital.
Human operator holds the kill switch and approves every change to the rules.

The protection stack

Every position carries a live stop (trailing or bracket). The protector audits coverage hourly — a naked position is a violation, not a choice.
Portfolio caps in code: max 25% per sector, max 20% in index ETFs, exposure bands set by market regime.
Adversarial review before every buy: an independent AI skeptic attacks each thesis (reward:risk ≥ 2:1 momentum / ≥ 3:1 reversion). A rejected trade is not taken.
Account circuit breakers: daily-loss breaker halts all buying; drawdown escalation tightens stops; a kill switch stops everything.
Time discipline: stale positions (10+ days going nowhere) are flagged and re-underwritten or exited.

Capital discipline

Cash never sleeps: idle cash is swept daily into 1–3 month T-bill equivalents earning money-market yield; buys redeem from the reserve automatically.
No trade is forced: when no setup clears the bar, the honest answer is zero trades — discipline over activity.
Regime-aware sizing: deployment targets follow measured market regime (trend, breadth, credit, volatility), not mood.
Market-closed guards: the fleet self-verifies the exchange calendar — no phantom activity on holidays.

Radical transparency

The complete record, resets included: the chart above chains every era of the fund — nothing is cherry-picked or restarted quietly.
Daily recap published after every close: trades done (or the honest “no trades”), protection coverage, NAV.
Un-edited audit trail: every figure on this page is generated by the system itself — no human touches the numbers.
Paper first, real capital gated: this is a forward test; real money waits until the weekly analyst’s go-live criteria are met.

Machine audit trail

Most recent system events, unedited

Loading…

Operator & legal framework

Who runs this, who decides, and what protects a future investor

Operator: Luxverum Capital is operated by ARMKU LLC, a Texas limited liability company. ARMKU sets the rules, holds the kill switch, and answers for the system.
Who decides: trade decisions are made autonomously by AI agents — but only inside human-defined, code-enforced constraints (mandatory stops, sector/ETF caps, circuit breakers, adversarial pre-trade review). Humans set the guardrails; the AI cannot change them.
Not financial advice: nothing on this site is investment advice, an offer, or a solicitation of securities. Luxverum Capital and ARMKU LLC are not registered investment advisers.

No client funds: this is a paper-trading forward test. No outside money is accepted, and none will be until the go-live criteria are met and proper legal documentation exists (e.g., applicable exemptions).
Investor-protection design: every safeguard is enforced in software and audited by independent AI roles; results are published unedited, resets included, with a downloadable monthly factsheet for independent comparison.
Hypothetical results: past performance, real or simulated, does not guarantee future results. Full notices: legal & disclosures.