A quant research harness for retail traders
Describe an idea in plain English. Lagias evolves it into a backtested strategy and a verdict on how much to trust it, including the odds it’s just overfit.
Waitlist open. Alpha invites start in August, launch follows in September.
| candidate | model | IS Sharpe | OOS Sharpe | verdict |
|---|---|---|---|---|
| Mean-Reversion Momentum | ridge | 1.19 | 2.05 | held up |
| Enhanced MR & Momentum | xgboost | 6.27 | 0.95 | fragile |
| Momentum & MR Hybrid | gbm | 5.73 | 1.30 | fragile |
| Revised Mean Reversion | lasso | 0.91 | 1.44 | held up |
| MR w/ Enhanced Features | gbm | 4.20 | 2.62 | held up |
| Volume Reversal Momentum | ridge | 0.77 | 0.12 | fragile |
A scripted walkthrough of the product we're building. All numbers are real research output. Backtests, not promises.
Research born at
Why now
None of them will tell you when not to trade one. 62% of US retail investors already use AI to inform investment decisions, yet only 23% mostly or completely trust its output (Investing.com survey, April 2026). We close that gap by grading every strategy on data it never saw, with a system the AI can’t flatter.
They drop you in a generic ETF mix and go quiet. No research, no control.
Every strategy looks like a winner. None of them tell you whether the edge is real.
A broker earns when you trade. A robo earns on your assets. Telling you to sit one out pays neither.
How it works
“Trend-following on tech, but I hate drawdowns.” Or just “build me something aggressive.”
The AI proposes generations of candidates. Each one is graded on data it never saw, and the strongest survive into the next round.
A straight verdict in plain English and the probability it’s overfit. You decide.
How it grades
Our AI invents and combines strategies, but it never scores its own work. Every candidate is graded by a deterministic harness using the same overfitting-aware statistics top quant funds use, and the strongest survive into the next generation. The approach is inspired by DeepMind’s research on evolving mathematical proofs.
Invents and mutates candidate strategies.
Deterministic, overfitting-aware scoring on data the model never saw.
The strongest candidates survive and are refined again.
A final verdict on how much to trust it, and why.
Why it matters
That’s what we’re building. Ask in plain English and get back a graded strategy with a trust score. No terminal, and no second-guessing whether the backtest is flattering you.
Example verdicts
The edge survives out-of-sample and through walk-forward. This one earns its place.
Sharpe deflates from 2.4 to 0.2 once we account for how many variants were tried. Don’t risk real money on it.
Same rigor, opposite answers. We have no reason to make either look better than it is. Figures are illustrative.
No conflicts, by design
We’ll charge a flat subscription, never a cut of your trades or your assets. So we can afford to tell you when the answer is “don’t.”
Two ways to use it
Feed it ideas, constraints, universes, and risk limits. Lagias becomes your personal research team. Go as deep as you like. Plain English by default, with the raw factor code and statistics underneath when you want them.
Say “aggressive” or “conservative” and let it propose. You get graded options with the risks spelled out, and you keep the veto on every one.
Plugs into your tools
Lagias is built for retail traders and the agents they use. Ask for a strategy in ChatGPT or Claude, and it’s built and graded by the same harness, with the same verdict, including “don’t trade this.” It connects over MCP, and the technical details live here.
Go deeper
Everything above rests on statistics you can check without taking our word for it. The handbook explains them in plain language, including the places where the honest answer is unhelpful. The blog argues with the industry, starting with our own product.
Early access
You’ll describe it in plain English and get back a backtested strategy with a verdict on how much to trust it. Join the waitlist for first access.
Alpha access opens to the waitlist in August 2026; launch follows in September. No payment details.
FAQ
No. Lagias is a research and education tool for self-directed investors. It doesn’t recommend specific securities to you personally. You make every decision.
Never. The product’s job is telling you how likely a strategy is to keep working, and often the answer is that it isn’t. Anyone promising returns is the thing we’re built against.
No. Describe your idea in plain English and read the plain-English verdict. The raw factor code and statistics are underneath if you want to dig. Go as deep as you want, or not at all.
Not yet. The research engine behind Lagias is real and working; the product around it is being built now. Alpha testers from the waitlist get access on a rolling basis starting August 2026, and launch follows in September. We would rather tell you that plainly than pretend otherwise.
Robo-advisors pick a generic mix and stay quiet. Build-your-own tools backtest a strategy once and show you the flattering curve. Lagias evolves generations of candidates and grades every one on data it never saw, and because we’ll charge a flat subscription with no cut of trades or assets, the verdict has nothing riding on it but being right.