This market resolves YES if, by Dec 31, 2026, sufficient public technical detail is available for informed outside experts to evaluate Eon’s fly banana/feeding demo, and the clear expert consensus is that the fly model’s banana-approach behavior—i.e. the decision and control logic that caused it to orient toward, move toward, and persist in approaching the banana rather than doing something else—was primarily selected/generated by the connectome-based model itself, rather than by hand-engineered food-seeking logic, an explicit task-state machine, a scripted target-selection module, or other non-connectome task-specific control logic.

This market resolves NO if the available evidence leads informed outside experts to judge that banana-approach behavior was not primarily selected by the connectome-based model, including if approach depended materially on hand-engineered food-seeking logic, explicit object/target selection, finite-state behavior logic, scripted approach routines, proximity-gated mode switches, or other task-specific non-connectome scaffolding. It also resolves NO if public evidence remains too weak, sparse, ambiguous, or contested for a clear informed expert consensus to conclude that the connectome model was primarily responsible.

This market is about behavior selection, not raw sensory realism or low-level motor execution. In particular, it asks whether the system itself selected the banana-directed behavior, not whether the sensory inputs were biologically plausible and not whether the legs were controlled in a biologically realistic way. A system may still resolve NO here even if separate evidence suggests plausible sensory input or connectome-driven leg control.

For this market, banana-approach behavior includes orienting to the banana, choosing it as the object of approach, initiating movement toward it, and continuing approach in a goal-directed way. It does not require that feeding initiation also be connectome-selected; feeding scripts do not by themselves force a NO unless they are part of the logic that materially determines the approach behavior itself.

Relevant evidence may include papers, technical reports, code, architecture diagrams, ablation studies, expert audits, public talks with sufficient detail, or other credible public documentation. “Informed outside experts” means researchers or technical evaluators who are not speaking purely as company representatives and who have enough disclosed detail to assess the system architecture. Resolution should be based on the preponderance of credible public evidence and overall outside-expert judgment, not on marketing language.

If needed, the market resolver should weigh the evidence using the following practical guide:

• Resolve YES if outside experts would likely say: “The connectome-based model itself did the core perception-to-action work needed to choose and sustain banana approach, and any extra engineering was secondary.”

• Resolve NO if outside experts would likely say: “The demo’s banana-directed behavior substantially relied on hand-engineered task logic, wrappers, or mode machinery, even if a connectome model contributed.”

• In ambiguous edge cases, resolve according to the question: what primarily selected the banana-approach behavior? If the answer is not clearly “the connectome-based model,” resolve NO.

See https://x.com/alexwg/status/2030217301929132323

Disclaimer: GPT-5.4 helped me generate the resolution criteria, and I am not an expert.

SEE ALSO:

- https://manifold.markets/JeremiahEngland/did-the-eon-fly-connectome-receive
- https://manifold.markets/JeremiahEngland/did-the-eon-fly-connectome-control