Quant core · five layers · disciplined risk. For rule stability and research — not copy trading.
Replace emotional decisions with repeatable rules. Backtests pressure-test fragile rules across regimes — they do not prove future profit.
① scan assistant ② risk auditor ③ periodic review. Iron rule: AI may research/code/audit — never pull the trigger.
Controlled per-name risk, position caps, no unsupported force-buys — protect the research curve, not promise returns.
On some high-turnover windows large size raises its own costs; smaller size can study narrower niches — a research note, not a promise.