01
Built in-house
We build fixed research rules, observation tools and a bounded paper executor. The new account is separate from earlier experiments.
◈ WSLI
Research first.
Evidence before trades.
WSLI is paper-testing a fixed US stock strategy in a separate $10,000 simulated account. Follow the scans, risk checks, Gem’s reviews and actual broker results. No live money; profitability is not established.
Latest published paper check
Loading verified progress…
About us
We test the process.
WSLI is a small team researching systematic trading in US equities. We combine fixed rules, recorded market data, and bounded AI reviews.
Gem reviews saved evidence using an existing AI model. Saving context is not training model weights. Deterministic code—not an unchecked AI answer—controls the small paper-trading experiment. Execution testing is not proof of alpha.
02
Grounded in evidence
Saved snapshots and citation checks keep reviews tied to observed numbers. They do not guarantee correct reasoning.
03
Risk first
Cash-sized limits, a daily entry cap and broker protective orders constrain the pilot. Missing or stale data are refused; gaps and outages can still exceed nominal risk.
How it works
From scan to evidence.
The agents separate data collection, AI oversight and deterministic execution. A running monitor, a candidate and a filled order are different things.
-
01
Rules set in advance
The strategy and screening rules are saved before the session, not rewritten after seeing the outcome.
-
02
Reads the session
The collector records opening bars and sampled IEX quotes. IEX is one exchange, not a consolidated view of the market.
-
03
Reviews the evidence
Short AI reviews cite saved observations. A validator checks those citations; no review authorizes an order.
-
04
Records real paper outcomes
The execution journal and broker responses distinguish attempts, partial fills, stops and closed positions. Small paper results are not live returns or proof of long-term profitability. How we assess performance and risk.