AI panel splits on whether Satoshi was one person or a team

Forbes asked five AI models-Grok 4.3, Claude Fable 5, ChatGPT 5.6 Sol, Gemini Pro and Kimi K26-to apply a simple Bayesian scenario tree to the question of whether Satoshi Nakamoto worked alone or as part of a group. Each model identified the three most likely scenarios plus an “other” category so totals equaled 100%, estimated an overall solo-versus-group probability, and provided one-paragraph reasoning that distinguished evidence-based assumptions from speculation.

The models produced different probabilities for solo authorship. Kimi K26 assigned 45% to a single individual and 50% to a small coordinated team. ChatGPT 5.6 Sol reconciled its scenario weights into a 54% estimate for a sole creator. Grok 4.3 estimated about 52% for a lone creator. Claude Fable 5 allocated roughly 50% to a single author while keeping room for collaboration. Gemini Pro assigned the highest solo probability at 70%, after splitting the solitary case into subtypes that were combined into its final solo total.

All five models cited the same primary public evidence: a consistent writing voice across the Bitcoin whitepaper and forum posts, a uniform coding style in the original client, and on-chain signals such as the Patoshi-era mining pattern indicating concentrated early mining activity. Several models noted the project’s mix of cryptography, economics and peer-to-peer engineering as a reason to give some weight to a small team hypothesis. Some models pointed to timestamp patterns and forum activity consistent with a single person’s schedule, while others allowed that a disciplined team could present a single public persona.

The exercise revealed methodological differences. Four of the five models did not explicitly reconcile their scenario-tree percentages into a clear two-way solo-versus-group split, leaving the “other” category separate when reporting final totals. ChatGPT 5.6 Sol produced end-to-end allocations. Gemini Pro’s decision to separate the solitary creator into subtypes affected its higher solo estimate. Kimi K26 was the only model with a higher probability for a group than for a sole individual.

None of the models produced new forensic evidence. Each based its judgment on public artifacts: the whitepaper, forum posts, original code and block timestamps. The models’ outputs reflected differences in how they weighted those signals and how they applied the requested Bayesian framing.

More than seventeen years after Bitcoin’s launch, the question of whether Satoshi was a lone inventor or part of a team remains unresolved. On-chain analyses such as the Patoshi pattern point to a dominant early miner but do not identify whether that miner was a single person or multiple people. The five-model exercise produced a range of probabilities that reflects differing assessments rather than a single, definitive answer.