Jacob Steeves

Jacob Steeves was billed by Proof of Talk Paris 2026 as Co-Founder of Bittensor. The archived programme lists Steeves for the fireside session “Decentralising Intelligence: The Unification of AI and Bitcoin” at 10:05 on 3 June 2026 and for the Bittensor-track Proof of Pitch session later that day.

The research record connects Jacob Steeves with work on decentralised machine learning. Steeves was an author of “BitTensor: A Peer-to-Peer Intelligence Market”, first submitted to arXiv on 9 March 2020. The preprint proposed a market in which machine-intelligence systems rank one another and higher-ranking peers receive rewards. The current arXiv record marks the paper as withdrawn, incomplete and obsolete, so it should not be described as a current or peer-reviewed foundation paper.

In 2022, Steeves was the first-listed author of “Incentivizing Intelligence: The Bittensor Approach”, whose authors were affiliated with the Opentensor Foundation. The paper appeared in the Trustworthy and Socially Responsible Machine Learning workshop co-located with NeurIPS 2022. It described Bittensor as a decentralised peer-to-peer machine-learning protocol with network-based evaluation and incentives.

Steeves, listed as Jacob Robert Steeves, was one of thirteen authors of the BTLM-3B-8K paper submitted on 20 September 2023. The paper describes an open-source language model with three billion parameters, trained on 627 billion tokens from SlimPajama using context lengths of 2,048 and 8,192 tokens.

Steeves was also one of six authors of “Incentivizing Permissionless Distributed Learning of LLMs”, submitted on 27 May 2025. That paper describes Gauntlet, an incentive system deployed on Bittensor and used in a permissionless distributed training run for a 1.2-billion-parameter language model.