Show HN: Belgi – deterministic acceptance pipeline for LLM outputs
github.comHello HN. I am releasing BELGI, a protocol + reference implementation for deterministic acceptance of LLM-produced artifacts (code/spec/docs).
Premise: LLM outputs are untrusted. BELGI only accepts outputs that can be verified against a locked spec + cross-file invariants, with a default NO-GO posture on unknowns or missing evidence.
Mental model: lock -> verify -> seal. Gates decide GO / NO-GO and produce replayable, tamper-evident artifacts.
I’d love feedback on:
where this fails in real CI pipelines,
missing invariant classes / bypass vectors,
whether default NO-GO is usable at scale (and how you’d stage it). Quick links: protocol repo (kernel + canonical specs): https://github.com/belgi-protocol/belgi canonicals/spec: https://github.com/belgi-protocol/belgi/blob/main/CANONICALS... whitepaper: https://github.com/belgi-protocol/belgi/blob/main/WHITEPAPER... org: https://github.com/belgi-protocol Demo video (X): https://x.com/belgiHQ/status/2013682594265661545 What BELGI does not do: it doesn’t make LLMs deterministic — it makes acceptance deterministic it doesn’t replace tests/review — it enforces evidence + invariants before GO