Harness²

All substantive input is from a single HackerNews thread (15 comments; vote counts are uniformly +80, so relative weight cannot be inferred). Reception is dominated by skepticism about novelty. One top commenter reduces the mechanism to: 'Your harness tells agent to do A when it encounters B. When agent encounters C and it doesn't know what to do it does D and you correct it to do E. Harness keeps track of this and makes agent do E next time it encounters C' — then others say this is 'something anyone developing AI systems already do. Learn from past runs to update your skills and system prompt... I have done this a lot over the last year+.' Another directly compares it to Hermes-agent ('harnesses, or at least Hermes-agent, are/is already doing' learning-from-feedback with reusable skills) and asks 'is there anything different in kind here going on than they're already doing?' A second theme is poor communication: the pitch 'plug-and-play framework for recursive agent harnessing in an open world, where agents encounter tasks unseen during harness design' drew a flat 'Huh?'; the official 'video overview' is mocked as 'just an animated slideshow of benchmarks. I guess that tells us who the intended audience is'; and Google is mocked for 'using slop images in readmes now.' The most concrete practical complaint: 'the lack of a single end to end example you can actually run' bugs one commenter more than the readme image. Trust in the publisher is also questioned: 'They had Gemini CLI as open-source, used the community for validation, then moved to closed-source Antigravity CLI. If the harness is closed, why open the harnessing?' A minority of comments are tangential (web ad-slop, YouTube content economics). One genuinely useful pointer: the readme badges link an arXiv paper (arxiv.org/pdf/2610.14224) that commenters found more helpful than the readme itself.

No aggregate ratings were found for this product during the last harvest.