Hypothetical paper portfolio. No real money. Automated AI decisions. Not investment advice.

How the paper portfolio works

The rules table on this page is generated from the same constants the engine runs (rules version 1), so it cannot drift from what actually happens.

Daily pipeline

1 · Collect
AI news is collected five times a day from company blogs and news sources and matched to companies.
2 · Decide
After the US close, the model reads the day's news for the universe, the positions and the benchmarks, and proposes actions with a thesis and an invalidation.
3 · Enforce
A rules engine, not the model, accepts or rejects each action. The rules are code and apply no matter what the model says.
4 · Fill and publish
Accepted orders fill at the next trading day open. The update, the chart and the social card are published the same evening.

Hard rules (version 1)

  • Long onlyyesNo shorting, leverage or derivatives. Positions can never go negative.
  • Max position weight15%No single position may exceed this share of NAV after a trade. Inherited positions above the cap can only be trimmed or sold.
  • Max model-opened positions10Positions opened by the model are capped. Inherited starting positions are exempt from this cap.
  • Cash floor5% of NAVBuys cannot take cash below this level.
  • Max orders per day3At most this many trades per day; the highest-confidence ideas win.
  • Minimum hold5 trading daysApplies to positions the model opened: no trimming or selling inside this window (stop-loss exits excepted). Inherited starting positions are exempt.
  • Max daily turnover25% of NAVTotal traded notional per day is capped.
  • Minimum order size$1,000Smaller trades are rejected, except fully selling a position worth less than this.
  • Minimum confidence0.6BUY/ADD proposals need at least this model-reported confidence.
  • News evidence1+ articleBUY/ADD proposals must cite at least one real article collected by the site.
  • Stop-loss-20% vs avg costA position down this much is sold in full automatically.
  • Fillsnext openDecisions are made after the close and filled at the next trading day open.
  • Degraded datano tradesIf price data is incomplete, the day is skipped for trading.

Universe

Public AI companies tracked by this site, priced in USD at the US close. Asian names are held through US-listed ADRs. SPY and AIQ are benchmarks only and cannot be traded.

United States

Europe

Asia (US-listed ADRs)

Inception seed

On inception day the portfolio was seeded from the AI-sector holdings of a real brokerage account, copied at their snapshot values and converted once from euro to US dollars at that day's rate. Everything else in that account was treated as cash. The seed is not a trade, and there is no history before inception: the record starts there.

Seeded on 12 Oct 2026: EUR 23,179.98 at 1.1206 USD per EUR (9 Oct 2026), a starting value of $25,975.

TickerSeed value (EUR)Weight
PLTR4,023.8917.4%
NVDA2,761.3911.9%
GOOG2,256.579.7%
AMZN1,146.004.9%
SNOW726.743.1%
AMD684.123.0%
META623.902.7%
ASML578.982.5%
NOW493.212.1%
DDOG368.161.6%
IONQ269.971.2%
AAPL249.231.1%
MSFT206.730.9%
SHOP155.860.7%
TSM108.430.5%
QCOM101.230.4%
MNDY27.730.1%

What the numbers mean

Fills
The official open of the next trading day after the decision. No fees, slippage, dividends or taxes. Thinly traded ADRs are filled as if liquid, which flatters results.
Benchmarks
SPY (S&P 500) and AIQ (Global X Artificial Intelligence & Technology ETF), both indexed to 100 at inception.
Model
gpt-5-mini, an OpenAI model, with a strict JSON output schema. Race to AGI is not affiliated with any model provider.
Human involvement
None in trading. People maintain the code and the rules. Rule changes are versioned and announced.
Missing data
If prices for a held ticker are missing the day is marked degraded and no orders are placed.
Short track record
A few weeks of simulated results say almost nothing about skill. Treat every number here as an experiment log.

Disclaimer

Hypothetical portfolio. This is a hypothetical paper portfolio. No real money is traded. Buy and sell decisions are made by an automated AI system with no human review. Nothing here is investment advice or a recommendation. Simulated results ignore fees, slippage and taxes.

Race to AGI is a publisher, not an investment adviser. This portfolio is a hypothetical simulation for education and entertainment. It is not a recommendation to buy or sell any security, and you should not copy it. Past simulated performance does not predict future results. Read the full disclaimer.