Updated · 10 Nov 2026 · 20:05 UTC

← All experiments

Public portfolio challenge · Research only

Public Portfolio Challenge

By Austin Starks · NexusTrade

Can an AI-designed strategy survive outside a backtest?

A documented portfolio challenge that makes strategy development and the experience of running it in public part of the story.

Creator-reported real-money accounts · Not independently verified

Active · Checked

The two separate challenge accounts publish observations dated 26 October 2026; the repository also has a 26 October scoreboard update. These are account observations, not proof that an LLM made a new trading decision that day. Lifecycle evidence ↗

Evidence & tracking status · Research only

TBT is not collecting this experiment. Research reviewed 15 October 2026.

Linked public account-value history is available. Reported percentages can be affected by deposits and withdrawals. The relationship between AI strategy authorship, deterministic execution, and human intervention still needs qualification.

Research only: establish material ongoing LLM involvement, cash-flow treatment, run boundaries, attribution, and the basis for permitted reuse. Account value alone is not a cash-flow-adjusted investment return.

Why follow this experiment?

  • Public writing documents the strategy and the choices behind it.
  • Different portfolios have different starts and mandates; they should retain separate identities.
  • AI-designed rules and an AI making ongoing trade decisions are distinct forms of involvement.

Building a continuous record.

Research only: establish material ongoing LLM involvement, cash-flow treatment, run boundaries, attribution, and the basis for permitted reuse. Account value alone is not a cash-flow-adjusted investment return.

Open the evidence disclosure →TBT’s source criteria →