# Novelty Diligence
## Candidate claim
For bounded service models, combine provenance-separated goal, phenotype, and
viability authorities with explicit hypotheses that may mark desired state faulty;
exclude each suspected authority from planning under that hypothesis; approve only
a transition independently shown viable across every surviving hypothesis; refuse
when no such transition exists.
## What is established
Feedback reconciliation, MAPE-K/autonomic loops, runtime architectural models,
runtime verification, invariant-constrained self-repair, N-version specification,
signed histories, rollback, and Byzantine fault techniques are established. None
is claimed.
## Diligence result
The preliminary sweep found strong collisions on every ingredient but did not find
an exact academic or product match for the full narrow protocol. That is a
research lead - not proof of legal novelty, patentability, freedom to operate, or
commercial uniqueness. The strongest non-claim is that runtime models plus
invariant checking already constrain adaptive repair.
## Claim discipline
Do not say Telosieve “knows true intent,” solves Byzantine repair generally, or is
the first self-healing system. A publishable claim requires a systematic search,
working implementation, baseline results, ablations, and independent review.
The 2026-07-29 post-M2 refresh is recorded in
[DILIGENCE_REFRESH_2026-07-29](DILIGENCE_REFRESH_2026-07-29.md). It does not
upgrade the candidate claim.