By Dec 31, 2026, will released technical details indicate that the fly connectome model in Eon’s demo received raw pixel input (or a close analog of photoreceptor-level visual input), rather than hand-engineered symbolic/per-object/per-task features?

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

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