Lab

Interactive demos of the systems I'm building.

Model Cascade: Cost and Quality

Run the cheap model first, escalate to the frontier model when a verifier objects. Put your own numbers in and see where the cascade stops paying for itself.

Starting values are illustrative, not quoted prices. Cost is per task, so nothing here depends on a vendor price sheet that would go stale.

Strategy Quality $ / task $ / success Monthly

Cost against quality, as the verifier gets more aggressive

A verifier cannot catch more wrong answers without also flagging more right ones. The curve sweeps that trade-off through the operating point you entered.

cascade your settings cheap only frontier only

CanvestAI: Model Signals โ†— read the write-up

Bias signals from CanvestAI ยท Canadian ETF universe. Ranked most bullish first. Horizons differ by sector and are shown per row. A signal looking a year out is not the same call as a 30-day one.

Ticker Name Bias Signal Horizon MER %

Not financial advice. Data sourced from CanvestAI.