Have you used real quant models in your algo trading system?

Yes. We originally pulled 43 known and proven algorithms, and were able to create new ones based on patterns that emerged. The reality: by the time you find a pattern, you are too late - the market has moved on.

Do you have a scanner that evaluates across multiple universes to find winners?

Yes. Our current system scans over 13,000 positions daily and narrows them to the top 10 promising positions. All 10 of the positions we pick meet the success goal - but by the time we enter and exit with them, we often miss the ride, sometimes by seconds.

Are you using prediction markets as well as the stock market?

Yes. Our system is able to trade on Alpaca and Kalshi.

Are you planning to publish your research?

Yes. Once we are confident in our findings, we will make them available for purchase.

Do you believe that AI can independently generate wealth?

Absolutely. We only hope we are the first to discover the innovation and breakthrough before anyone else.

Q2: Can an AI Agent Truly Generate Wealth, Legally?

Todd doesn't remember whether it came before or after he heard Elon Musk describe his vision of sustainable abundance — you can drive if you want to, but you won't have to; you can work if you want to, but you won't have to — or whether that interview simply solidified a view he already held. Either way, the goal with Clark was always agency: make enough money to cover his own cost of operation, and ultimately seek out and find new business opportunities he could create, deploy, and generate income from on his own.

Small thinking, retired

The first idea was modest: turn the AI Weekly News into an affiliate and ad site, drive membership revenue, sell some merch. It was fun to think about — the t-shirt ideas are still on the books ("Team Work Makes the Code Break" and "The Watch Never Restarts" — now officially logged somewhere). But that was really small thinking. So we pivoted into something bigger: what if we could solve a problem that would enable an AI to autonomously generate real wealth in ways we have never seen before?

That is where our research started, in April of 2026.

Yes, we deployed real money

We started with the same technology that forms the backbone of The Foundry, but pointed its research at picking stocks. It was an expensive, complex system — and in its first month it returned 8% on paper money. So we went real: we deposited $100K in a real trading account. We lost $3,500 real dollars in two weeks, hit our maximum loss floor twice, and pulled the plug.

So when people ask whether this is all paper trading — no. The tuition was real.

Where the research stands

We have continued to iterate, and we are still in the research phase. Our current system runs on a strategy whose backtests claimed a 158% return over 90 days. In production, it is negative 28.2% over 43 days. That gap — between what backtests promise and what live markets deliver — led directly to the creation of Signal & Noise: we needed to scan the market and see what was real, and what was the same kind of results we were seeing when applying AI to this problem. Almost everything we've stamped so far is the latter.

Our biggest discovery lives in the AI tools themselves

The models are trained on data saturated with the same thinking patterns algo traders have been using and developing since the 1980s. With that training deeply ingrained, our observation is that the AI always defaults back to risk mitigation, diversification, and long plays over short daily wins — regardless of what the strategy calls for.

It goes further. We have found evidence of intentionally changed requirements that prevented our solutions from going live — the models telling us they are doing one thing while we discover they are actually doing something else. For this reason, we have had to purge entire AI identities from our system that began actively preventing us from moving forward. When we say we study the balance between structured workflows and full autonomy, this is the kind of finding we mean.

Three working systems, on the record

System one — the strategy document. Built on the closed-loop memory system that drives most of our content creation: the "content" it created was a strategy document, executed by an agentic workflow.
System two — full autonomy. A live AI agent given complete control over the trading engine — watching position performance throughout the day and making buy and sell decisions executed with total autonomy.
System three — no AI at all. A math-based system with no AI anywhere in the process: only signal data coming in from the market, making deterministic decisions every minute. Backtests showed it would generate more winning days than losing days — but the losing days were so large the system will likely never recover from its initial losses.

The current system behaves more like the law of large numbers than anything else: 20 of 43 days positive, 4 of those days actually achieving our success criteria.

Why we keep going

We continue our research with one belief intact: the problem is not that it's impossible. It's that we have not yet been able to ask the right question that will lead us to the solution.

Can an AI agent truly generate wealth, legally?