Portfolio Performance

$1M Equivalent Portfolio

—

normalized from live return

Total Return Since Inception

—

since Feb 17, 2026

Sharpe Ratio

—

risk-adjusted return

Max Drawdown

—

on realized closed trades ·

Proprietary Technology

Deliver differentiated alpha with AI-powered quant technology built for disciplined execution.

Obsidian Quant evaluates independent market signals across price action, volume, momentum, and execution context before capital is deployed. Each trade cycle is logged, reviewed, and fed back into the system's critique layer, allowing the model to improve its decision process while maintaining strict risk controls.

Scale view: Portfolio equity is normalized to a $1,000,000 starting portfolio using the same live return profile. Actual account custody and deployment size may vary by client.

Cumulative P&L

All closed trades, inception to date

—

Risk & Return Metrics

Sortino Ratio

—

downside-adjusted

Total Trades

—

— closed trades

Profit Factor

—

gross win / gross loss

Beta to S&P 500

—

market sensitivity

Performance vs. Major Indexes

	Trailing 5 Sessions	Since Inception
Loading benchmark data…

Index figures are price returns of the SPY, QQQ, and IWM ETFs measured over the identical windows as the portfolio. Portfolio return is measured on live account equity, including unrealized P&L.

Institutional Deployment

Capacity is deliberately limited — and shared across every licensee

Capacity

Limited

shared across all licensees, not per account

Leverage

~1.2×

modest by design; Reg T permits 2× overnight

Capacity is finite, and that is the point. The opportunities this system takes are too small for institutional capital to compete for — which is precisely why the edge persists. Capacity is shared across every licensee, not allotted per account: all accounts trade the same signals at the same moment, so the binding constraint is the total book rather than your balance. We have measured that ceiling and we allocate against it. Seats are limited and offered by invitation to qualified parties.

Who we can take. Seats go to parties who can deploy at least $250,000 in a margin-enabled brokerage account they own and control, and who are comfortable with a strategy that wins a minority of its trades and is expected to have losing stretches. We do not take custody, and we cannot move money out of your account. If that fits, tell us your situation and we will come back with availability and terms.

Request a Seat

Proprietary Technology

Small losses by design. Winners left to run.

Obsidian Quant is a systematic long-only US equity system. Every position enters with a predefined maximum loss and no ceiling on its gain — the average winner is several times the average loser, which is how the system compounds while winning a minority of its trades. Execution is deterministic: a fixed, tested rulebook applied identically to every position, with no discretion at any point. The research around it never stops — hypotheses are generated and tested across multiple market regimes autonomously, on criteria fixed before each test runs. Most are rejected. Only a human promotes anything to live capital.

Multi-Dimensional Signal Fusion

Rather than relying on any single indicator, the system simultaneously evaluates multiple independent market signals across price action, volume dynamics, and structural momentum. Positions are only initiated when signals reach a statistically significant confluence threshold — eliminating low-probability noise trades entirely.

Recursive Learning Loop

Every trade the system takes becomes evidence about the system itself. Results are analyzed against the full historical record — by the founder working alongside frontier AI models — to surface systematic patterns, signal failures and emerging edge conditions. That analysis produces hypotheses, and those hypotheses are now tested autonomously: each one is registered with its pass criteria before it runs, evaluated across multiple market regimes including a bear market, and rejected on those criteria without human involvement. Results are reported against the number of hypotheses tried, so a survivor is never read without its denominator. The trading engine remains deterministic and does not rewrite its own logic — the loop can reject on its own; only a human can promote anything to live capital.

Shadow-Validated Strategy Evolution

Before any refinement touches capital, it must earn its way in. Candidate improvements are backtested across multiple multi-year market regimes — including full bear markets — then run as a parallel shadow portfolio against live market data, measured head-to-head against the production strategy. Only enhancements that demonstrate durable statistical edge are promoted, one change at a time, so that every result stays attributable to a specific decision. This is versioned software: the release running today is the accumulation of every change that earned its place, and the strategies that did not earn theirs were retired on the evidence. The system evolves, but never on faith.

Risk Architecture

Strict concentration caps at both the single-name and aggregate-portfolio level — no one position can dominate the book
Trailing stop levels tighten automatically as positions move into profit — open gains convert to protected gains, and stops never retreat
Protective orders rest at the broker, not in software — positions stay covered between cycles, and remain protected even if our infrastructure goes down
Asymmetric by construction: losses are capped by a fixed stop while gains are left uncapped — the asymmetry comes from letting winners run, not from a preset target
Hard per-trade loss limits enforced at the engine level — no single position can meaningfully impair the portfolio
Positions are reconciled against broker records on every cycle — stale or untracked exposure is closed automatically, never ignored

Execution & Coverage

Continuously evaluates a dynamic universe of several thousand U.S. equities throughout every trading session
Full scan-to-execution cycle completes every few minutes during market hours, with automated position management on every pass
All execution routed through institutional-grade API infrastructure with bracket order management
Liquidity screens exclude thin, hard-to-borrow, and illiquid names before capital is ever committed
Automatic API retry and reconciliation logic — transient broker or data failures degrade safely, never blindly

"The edge is not in any single signal. It is in the system's ability to recognize when multiple independent data streams converge — and to have the discipline to act only then, and nowhere else."

— Anthony W. Licausi, Founder

Next Step

Want to see this architecture running on live market data?

We walk through the signal stack, the risk controls, and the live track record on a short call — including how the technology is licensed so you keep custody and control of your own capital.

Request a Seat

Founder-Led Technology

Built by a practitioner at the intersection of markets, code, and execution.

Obsidian Quant was developed by Anthony W. Licausi, founder of Barenberg Capital and architect of DEALITHIC. It is the product of nearly three decades of coding experience, two decades of trading experience, and a career spent close to capital markets, investment banking, M&A advisory, underwriting, and deal structuring.

Nearly 30 Years Coding

Started writing code at age 10 and kept compounding that technical base long before AI became mainstream.

Years Trading

Built around real market experience, not abstract feature design or theoretical signal research.

Months Building

The hardest technical undertaking of Anthony's life: a live AI trading system with analytics, execution, critique, and risk controls.

Why This Exists

Anthony's background has always lived between finance and software: writing code from a young age, then applying that engineering mindset across trading, investment banking, LBO underwriting, sell-side advisory, capital raises, and transaction structuring.

That combination shaped the core belief behind Obsidian Quant: edge does not come from one magic indicator. It comes from infrastructure that can process independent data streams, enforce discipline, manage risk, and keep improving after the trade is over.

What Obsidian Represents

Obsidian Quant was built as more than a trading script. It is a recursive operating system for market decisions: screening, signal confluence, execution routing, portfolio analytics, and AI-driven critique working as one stack.

After 18 months of development, the platform is now being shared with individuals and institutions that want direct access to cutting-edge trading technology while maintaining custody and control of their own capital.

Builder's Arc

Age 10

First line of code written; the foundation for a lifelong systems-building obsession.

Mid 2000s

Entered investment banking and M&A advisory, working through live transactions and institutional underwriting pressure.

2025-2026

Spent 18 months building Obsidian Quant into a live recursive AI trading system.

Today

Opening access to the technology for qualified individuals, family offices, and institutions.

Talk To The Builder

Questions are answered by the person who wrote the code.

No sales team and no intermediaries — intro calls are taken by Anthony directly. Bring the hard questions about the architecture, the risk framework, or the track record.

Request a Seat

Recent Trade Activity

Closed Trades — Last 30 Entry/exit prices and signal detail omitted for confidentiality

Date	Instrument	Direction	Result	P&L
Loading trade history…

Obsidian Quant — Technology Licensing

Deploy our algorithms.
Keep full control.

You're not sending money to a manager. You're licensing a proprietary AI trading system and running it inside your own brokerage account. Your capital stays in your account at all times — we never touch it, never custody it, never have access to it. Think of it as deploying institutional-grade trading software to your own infrastructure.

HOW IT WORKS

Step 01

Open an Alpaca account & fund it

Create a brokerage account at alpaca.markets — it takes about 10 minutes. Fund it with any amount you'd like to deploy. Alpaca offers both paper (simulated) and live trading, so you can run the system in simulation first to validate performance before going live.

Step 02

We configure & deploy your instance

You provide your Alpaca API keys. We stand up a private instance of the system on a dedicated server — your own isolated environment, your own trade log, your own configuration. No shared infrastructure. No pooled accounts. Fully isolated.

Step 03

The AI begins trading. You keep control.

Every position, fill, and balance is visible in your own Alpaca account in real time, and performance summaries are delivered to you by email. You can halt trading or close positions at any moment directly in Alpaca — it is your account and your keys, so control never leaves your hands. Every account runs the same tested configuration — changes are validated across the entire book before they reach anyone, never tuned per client.

Common Questions

Is this a fund? Are you managing my money?

No. You retain full custody of your capital at all times inside your own Alpaca account. We license software. We have no access to your funds, no discretionary authority, and no ability to withdraw or move capital.

Do I need trading experience?

No. The system is fully automated and every licensee runs the same configuration. You can watch every position and fill in your own Alpaca account — the AI handles execution, position sizing, and risk management.

How is the fee structured?

Obsidian Quant is licensed as software for a flat monthly fee. There are no performance fees, no profit sharing, and no pooled capital arrangement.

Can I pause or stop trading at any time?

Yes — at any time, directly through your Alpaca account. You always have full control. Your subscription can be cancelled with 30 days notice.

TECHNOLOGY LICENSE

Obsidian Quant License

From ^$1,500 / month

Flat monthly software fee. Private deployment on dedicated infrastructure, running against your own brokerage account. No performance fee, no profit share, and we never take custody.

Capital deployed	Monthly fee
$250,000 – $1M	$1,500
$1M – $3M	$3,500
$3M – $10M	$7,500
$10M+	By arrangement

Your tier is determined by the capital you deposit, assessed at each monthly invoice. If deposits cross a threshold the new tier applies from the next invoice. Trading gains and losses never change it — this is a licence fee for software, not a fee on assets, and it does not float with your account balance.

Included with the license

Technology license
Private isolated instance — your keys, your account
Dedicated infrastructure — your own server, never shared
Onboarding call
Performance summaries delivered by email
Phone and email support

I'm Interested! Let's Talk

Technology License Disclosure. Obsidian Quant, developed by Barenberg Capital, licenses proprietary algorithmic trading software. We do not manage client assets, provide investment advice, or operate as a registered investment advisor (RIA). Clients retain full custody of their capital at all times within their own brokerage accounts. All trading activity is executed by the client's own account using automated software. Past performance of the system — including all results shown in this portal — is not indicative of future results. Trading involves substantial risk of loss. Licensing this technology does not guarantee profitability.

Is this trading real money?

Not yet. Every figure on this site was produced by a live-market paper account funded at $100,000 on 17 February 2026 — real market data, real prices, broker-simulated fills. No client capital has been deployed to date.

The founder did not arrive at this from theory. He traded his own capital discretionarily and made a living doing it. That experience, not a backtest, is where the risk architecture comes from — and it made the next step obvious: build the discipline into software, so results would no longer depend on the operator. The failures that end discretionary traders are rarely analytical. They are operator failures — and they are precisely what a system does not do. It does not forget a position, does not trade differently after a loss, and does not change position size because the rent is due.

What costs are missing from these returns?

Two, and we would rather name them than have you find them. Financing: the simulated account is charged no margin interest, and over this period the effect is under one percentage point. Market impact: the broker's simulator fills at the quoted price regardless of size. Execution slippage itself is measured rather than assumed — 0.07% per leg across 2,123 fills, benchmarked against the volume-weighted average price of the minute each order executed in. Our own backtests charge nearly double that.

Why are there two different maximum drawdown figures?

They measure different things, and both are disclosed.

On realized closed trades: 15.44% — and roughly four percentage points of that is a single day. Through 18 August 2026 the figure stood at approximately 10%. On 19 August, orders generated while the market was closed were accepted and queued by the broker rather than rejected. A queued order creates no position, so our reconciliation logic read that absence as a completed trade, cleared its own record, and allowed the same order to be placed again. They filled together at the next open as one oversized position, which was closed the same day at a realized loss of $17,678.

The defect was in order handling — not in how positions are selected, and not a market call. Orders are now refused whenever the exchange is closed, refused again if an order is already working in the same security, and refused again if the resulting position would exceed a fixed share of account equity. Each check fails to the safe side, none of them can block an exit, and a regression suite reproduces the original conditions to confirm the sequence cannot run again. No client account was affected.

Against daily account equity: 17.1%, reached in early June 2026 during a separate defect in protective stop-order handling that was identified and corrected. Between that correction and 19 August 2026 the equity-based figure was 4.1%.

Both incidents were defects in order handling rather than in position selection, both were corrected, and both are set out in the disclosures at the foot of this page.

How much capital can the system handle?

Less than most people expect, and that is the point. Returns depend on positions that cannot absorb unlimited size, and because every licensee trades the same signals at the same moment, capacity is shared across all of them rather than allotted per account. It is finite, we know what it is, and it is allocated accordingly. The edge persists precisely because the opportunity is too small for large capital to compete for.

What am I actually buying?

A software licence, for a flat monthly fee — not a fund, not a managed account, no performance fee. The system trades in your own brokerage account, in your name. We never take custody and cannot move money out of your account. Access runs through an API key you issue and can revoke at any moment.

Is there a minimum account size?

$25,000. It is a fidelity requirement, not a regulatory one. The system holds a diversified book of positions at once and buys whole shares — so below that level the individual positions get too small to hold the higher-priced names, and the account quietly ends up trading a cheaper subset of the strategy instead of the strategy. At $25,000 and up, every name the system selects is reachable. It also requires a margin account; a cash account cannot run it. We would rather decline a $5,000 account than have someone run a diluted version and judge the system on the result.

That is the technical floor, not the commercial one. The licence is a fixed monthly fee, so on a small account it becomes a meaningful annual drag before the system has done anything. In practice we begin conversations at $250,000 of deployable capital, below which the economics do not work for the client — which is the only reason that matters.

A minority of trades are winners. How does that make money?

By design. Losses are capped by a fixed stop; gains are deliberately left uncapped. The average winner is several times the average loser. Expect stretches of consecutive losing trades — that is the shape of this return profile, not a malfunction.

Has the strategy changed over the track record?

Yes. Improvements are tested across multiple multi-year market regimes and promoted one change at a time, so every result stays attributable to a specific decision. The consequence worth knowing: the record spans more than one configuration, and the release running today is newer than much of the history behind it.

What would cause this to stop working?

Overnight gaps, which no stop can protect against. A broad, sudden decline, since the book is long-only and modestly levered. Execution costs rising materially above what we have measured. Or a market with too little dispersion to work with. The record is also under six months and spans a generally rising market.

Who is behind it, and what happens if you disappear?

Obsidian Quant is developed by Barenberg Capital and led day to day by its founding principal — but it does not depend on him. Two additional principals are designated to step in, covering client relationships and technical operations between them, so the system keeps running and clients keep being served. Two structural protections sit underneath that: protective orders rest at the broker, not in our software, so open positions stay covered even if our infrastructure fails; and because you hold custody in your own account, you can never be locked in.

Still Have Questions

Ask them directly.

Anything not answered here gets answered on a call — including the parts we have not published.

Important Disclosures. This report is provided for informational purposes only and is intended solely for the authorized recipient identified by the access link. Obsidian Quant is a proprietary algorithmic trading system developed by Barenberg Capital. Past performance is not indicative of future results. All performance shown reflects paper trading activity on simulated infrastructure. Trading involves substantial risk of loss. This document does not constitute an offer to invest or solicit capital. Returns are presented on a cumulative basis over the stated period and are not annualized, consistent with the convention that performance for periods of less than one year should not be extrapolated to a full-year figure. Beta and correlation are measured against daily SPY returns over the same period. *Max drawdown is calculated from the realized closed-trade P&L series, consistent with the trade statistics on this page. Approximately four percentage points of the figure shown are attributable to a single trading day, 19 August 2026, on which a defect in order handling caused a series of duplicate orders to be generated while the exchange was closed; the orders were queued by the broker rather than rejected, filled together at the following open as one oversized position, and that position was closed the same day at a realized loss of $17,678. Prior to that day the closed-trade figure stood at approximately 10%. The defect lay in order submission and reconciliation rather than in the selection of positions; it was corrected the same day with checks that refuse any order while the exchange is closed, refuse a duplicate order in a security that already has one working, and refuse any order that would take a single security beyond a fixed proportion of account equity, together with a regression test that reproduces the original conditions. No client account was affected. Separately, measured against daily marked-to-market account equity, peak-to-trough drawdown reached 17.1% in early June 2026, during a development-phase incident in which a defect in bracket stop-order handling left open positions without functioning protective stops — that defect was identified and corrected in early June 2026, and between the June correction and 19 August 2026, equity-based peak-to-trough drawdown was 4.1%. Founder background information is provided for context regarding the developer of Obsidian Quant and Barenberg Capital's technology and should not be interpreted as investment advice, a guarantee of results, or a representation that future performance will match historical performance.