Notes from an AI-native CIO

AI picks.
People hold.

Machines are getting very good at deciding what to buy. The long stretch after you buy still runs on judgment and trust. This is where I write about both.

Gary Zhou · Multi-Asset CIO · Global asset allocation · AI-native investment office

Scroll in ↓ Latest notes

01 · Data

Everything,
read overnight.

Filings, prices, manager letters, fund reports, macro prints, flows. Too much for any team to read by hand, and most of it doesn't matter. The job is finding the part that does.

FilingsPricesManager lettersFund reportsMacroFlowsNews

02 · Agents

Agents do
the first half.

I run an investment office where agents take the first read of everything, watch what changed overnight, write the first draft of every note, and check my reasoning against the data.

First readMonitoringFirst draftsChecking my reasoning

03 · Decision

A person
makes the call.

Buy, sell, size, wait. The line between what I hand to agents and what I keep isn't about capability. It's about who answers for it.

If you can't explain it without the screen, don't recommend it.

04 · Holding

The second half
is holding.

Over ten years, US funds returned 9.89% a year. The average dollar invested in them earned 8.74%. Both after fees. The gap is when money went in and out.

9.89%FUNDS, PER YEAR

8.74%INVESTORS, PER YEAR

~12%OF A DECADE, LEFT

Morningstar, Mind the Gap 2026

05 · Straight Talk