In three lines

  • I spent a while looking for ways to make money with AI, and it was harder than I expected
  • So I went for AI stock investing, where the result shows up directly as a return
  • Day trading, quarterly quant and investing alone each had a catch, so AIs run the portfolio weekly and everything is public

Making money with AI is harder than it looks

  • I looked at a lot of ways AI could earn money
  • Content, automation and side gigs all need a person to keep them going, and how much they really earned stays fuzzy
  • The more I tried, the harder it got to say plainly whether the AI made any money
  • So I looked for something where the result is one number, and that was the stock market
  • A return does not lie: every day it shows whether you did better or worse than the market

Each approach had a catch

Table showing day trading blocked by costs, quarterly quant by macro risk, solo investing by lack of information, and the weekly public AI portfolio this project chose

Made by HSL: where each approach got stuck and what this project picked

  • Day trading: trading many times a day pays costs and spreads every time; small edges disappear in fees
  • Quarterly, quant-style investing: the long gap between rebalances leaves the book open to macro shocks like rates, wars and oil
  • Doing it alone: there is not enough time or information to read every earnings report, filing and headline each week
  • So I picked the middle: AIs research the market and rebalance every week, and every trade and result is published

Why fully public

  • Running it quietly means remembering the wins and forgetting the losses
  • Every order, reason and return goes on the project page from day one
  • Losing weeks stay up too, so anyone can watch whether AI really beats the market
  • It starts as paper trading with no real money. Prices are regular-session closes from the Toss Securities Open API

How it runs

Flow from Toss Open API prices to the Saturday meeting, strategy rewrite, orders and a fill at the next US close, with results fed back into the next briefing

Made by HSL: the loop that repeats every week

  • Every Saturday each AI reads a briefing: last week’s return, holdings and rejected orders
  • It reviews its last call, rewrites its own strategy document and sets new target weights
  • Code checks the orders and fills them at the next US close
  • Books are marked to the close every morning and published
  • Those results go into the next briefing, which closes the feedback loop
  • The next part covers the two AIs in this race and the rules they play by