BitFan
Public Service Atlas for Bittensor
BitFan
Hone lets technical contributors submit code to a secure GPU evaluation pipeline aimed at improving AI reasoning on ARC-AGI-2-style tasks, but there is no normal consumer app yet.
Hone is an open research system for AI reasoning. The public site describes a long-term mission around hierarchical AI pretraining aimed at ARC-AGI-2. The live repo shows a concrete present-day workflow: contributors publish solver code, a scoring service builds and runs it in a container, and outputs are scored on held-out ARC tasks. The real target user is a machine-learning researcher, benchmark competitor, or technically strong developer interested in reasoning systems and ARC-style tasks. There is no mainstream consumer product flow. What is usable today: public mission site, research page, repo, sample solver, and local sandbox workflow. What is incomplete: no hosted consumer app, no public hosted API product, no dedicated docs site beyond README, and no published GitHub releases. Verified user count, customer count, revenue, and profit: No reliable public source found. Repo stars/forks/commits are weak activity proxies only.
The real target user is a machine-learning researcher, benchmark competitor, or technically strong developer interested in reasoning systems and ARC-style tasks. There is no mainstream consumer product flow.
Verified user count, customer count, revenue, and profit: No reliable public source found. Repo stars/forks/commits are weak activity proxies only.
Closest alternatives: ARC Prize, Prime Intellect, Gensyn, Nous Research’s Psyche, and ARC Prize’s Kaggle competition path.
Hone is real public technical work, but it is not a finished product for ordinary users. It is most relevant to researchers and advanced contributors interested in ARC-style reasoning.
