Updated · 10 Nov 2026 · 20:05 UTC

← All experiments

Creator experiment · Research only

Beebots

By Mike Russell · Creator Magic / CreatorMagicAI

What happened when three trading personalities shared the same market?

Three bees, different instructions, and an official experiment that has now ended. Its setup changes and creator-reported conclusion remain a historical case study.

Creator-reported real-money trading · Not independently verified

ENDED · Checked

The creator announced the official experiment’s conclusion on 26 October 2026. Independent self-hosted installations are separate runs. Lifecycle evidence ↗

Evidence & tracking status · Research only

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

Earlier deployment research recorded public account snapshots and equity history reported as real money. TBT has not independently verified those balances. The repository defaults to paper trading; deployment and repository versions must be distinguished.

Before tracking: establish run identity, actual LLM versus fallback decision attribution, costs and funding, resets and cash flows, sufficient history, and a documented basis for collection and reuse.

Three yellow robotic bees exploring a cobalt alpine field
Editorial illustration · Not a trading record

Original experiment ended

The creator’s reported conclusion.

The creator reports $999 starting capital and $785.16 final capital: a 21.4% trading loss. Model calls cost another $42.35, reported separately from the trading account result. TBT has not independently audited these financial results.

Read the creator’s conclusion ↗

Community Hive and software

The official Hive is closed.

The Hive was a separate paper-trading experiment using simulated money. Its leaderboard is now closed; the repository remains available but is no longer maintained. Self-hosted paper installations can continue independently and do not extend the original experiment’s record.

Why follow this experiment?

  • Different instructions let the bees approach one market in different ways.
  • Model changes and deterministic fallback behavior are part of the experiment.
  • Preserved setup information and the creator’s published conclusion make this a historical case study.

What the record establishes.

The creator’s conclusion closes the official experiment; it does not independently verify every balance or decision. Earlier deployment evidence remains below. Any self-hosted installation would need its own run identity, decision attribution, cost and cash-flow evidence, and permission review.

Open the evidence disclosure →TBT’s source criteria →