Public Evolution is not "self-merging AI". The product contract is stricter: opt in, propose a bounded Draft PR, stay paused while it is open, help review other work, and resume only after maintainers merge or close it.
The user explicitly consents to public contribution. The machine is never supposed to start shipping public proposals without that opt-in gate.
The proposal is review-gated on purpose. Maintainers can inspect it, comment, merge it, or close it. The machine never auto-merges its own public work.
The machine should not stack endless PRs. It waits while its own Draft PR is unresolved and reuses cycles for scoped peer review of other proposals.
Merge or close the Draft PR and the machine becomes active again. That is the real pause/resume contract and the part that makes the system maintainable.
It avoids the category mistake of presenting self-improvement as autonomous self-merge. The human maintainer remains the final decision-maker.
While a proposal is open, other opt-in machines can help with peer review. That creates a more realistic contribution loop than just opening more PRs.
This is one of the rarest credible differentiators in NEXO, so it deserves a short public explanation instead of being buried in implementation details.
The product-facing explanation lives here. The deeper operational rules still live in the public contribution docs.