No autonomous AI deciding trades. The LLM appears once as an advisory grading gate. Everything else is rule-based and reversible.
About
A solo build out of Toronto.
Aegus is one operator and one reasoning layer, focused on a single product right now: PATH. Other products are shelved — not deleted — until PATH proves itself live. When they return, they'll route through the same Institute grading framework that powers PATH today.
how we build
Honest defaults.
We pull from SQL, files, and cached snapshots into prompts. No vector embeddings, no RAG. Same shape of evidence every time.
Every new strategy runs in paper for a measured cohort before live. We publish the refund-criteria bar and stick to it.
Code lives on GitHub. Decisions, kills, throttles, and post-mortems all logged. No black boxes.
PATH is free. Sign in with a Solana wallet — nothing custodial, no tiers, no founder seats. You stay in control of every key.
One operator, multiple AI specialists. Built from a Mac mini and a laptop, both in Toronto. Real cost, real edge.