Loading the latest verified snapshot…
Buy good businesses
Looks for understandable businesses with strong cash generation, manageable debt and a sensible price.
The AI interpretation can wait when nothing looks attractive.
A BUSY DAD. A REAL AI EXPERIMENT.
I hired three
AI investment legends.
Can AI investing strategies beat the market? I tested styles inspired by Buffett, Dalio and Soros—and shared what happened.
Three investing styles. Three AI setups per style. Each portfolio starts with US$10,000 in simulated money.
What are the three setups?
GPT 6 Astra and Qwen 3.8 27B Local are the AI models making the decisions. Each style runs with Astra once a day, Qwen once a day, and Qwen hourly.
DID THEY BEAT THE MARKET?
Loading the final comparison…
S&P 500 reference —Simulated SPY return over the same period
Loading the final results…
Loading the latest verified snapshot…
Read the economy
Builds a view of growth, inflation and interest rates, then allocates across broad stock, Treasury, gold and cash exposures.
This is a tactical macro interpretation, not a replica of All Weather.
Loading the latest verified snapshot…
Spot changing expectations
Looks for a concrete event that could change market expectations, then considers a focused position around that idea.
The AI interpretation states what would prove the idea wrong.
Simulated results, not income. These AI interpretations are not the actual investors or their endorsements.
HOW IT UNFOLDED
Day by day, since launch.
Return since start at each US market close, against the S&P 500 reference.
Loading the day-by-day chart…
Each portfolio started with US$10,000 of simulated money. Hover, tap, or use the arrow keys to read a day. Each style has its own scale, so compare lines within a chart rather than between charts.
View the numbers
RECORDED ACTIVITY
What happened?
The latest action from each investing style.
Checking activity…
No activity loaded yet.
Trades use simulated money. An AI decision describes its plan; it does not confirm a trade.
FOLLOW THE EXPERIMENT
Wins, losses and the days nothing happens.
The experiment is over. Follow for the wrap-up and what comes next.
ABOUT THE EXPERIMENT
How the experiment works.
Nine independent portfolios let us compare investing styles, AI models and how often they make decisions. They may hold overnight, keep cash or make no trade.
Astra · Daily
One decision at 10 a.m. New York time on each market day.
Qwen · Daily
One Qwen decision at 10 a.m. New York time on each market day.
Qwen · Hourly
Qwen checks at 10 a.m. and hourly through 3 p.m. on regular market days.
This short trial cannot prove that an AI style will reliably outperform the market.
Calculations and limitations
“vs S&P 500” is the portfolio return minus the separate SPY account return, in percentage points. It can be positive while the portfolio still loses money.
The final results and the day-by-day chart use each portfolio’s recorded equity at the 4 p.m. New York close on every trading day, with SPY taken from the closing reports. Values come from the exported reporting snapshot at its stated cutoff. A market-close snapshot is historical; an intraday snapshot is provisional. A fresh feed heartbeat does not make an old return current.
Returns and benchmark differences are rounded to two decimal places. Missing values remain unavailable and are never treated as zero.
Results use recorded cash and positions, observed fees and identified dividend adjustments, with modeled retail costs where available. AI, data and hosting costs are excluded.
Market-close marks use available IEX closing-minute bars, not an official consolidated closing auction. Missing activity, prices or distributions can change a provisional result. The separate SPY account is a simulated ETF reference, not an asserted official index total return.
Technical health and feed details
Checking the sanitized feed status…
This secondary monitor reports sanitized service heartbeat and durable ledger activity. It does not expose account identifiers, raw model packets or credentials.