EVIDENCE
Claims should have a failure mode.
We separate observation, reproduction, external confirmation and commercial proof. Two independent maintainers have confirmed lifecycle defects identified through our current reliability methodology. Neither relationship is presented as a paying customer.
CONTINUITY v1.4 · EXTERNAL CONFIRMATION
Ambiguity bypassed a frozen-constraint guard.
A cold-session evaluation began from a passing native baseline of 186/186 tests, then introduced an adversarial ambiguity: two frozen claims shared a substring and a new conflicting operation referenced that ambiguous substring.
The maintainer independently reproduced the ambiguity, described the diagnosis as correct, shipped v1.5.0 fixes, and added acceptance tests. The two provenance cases were also confirmed as real. Euphoric Doom did not independently validate v1.5.0 and does not claim that it did.
Source evidence: direct maintainer correspondence retained internally. This is an independent technical evaluation, not a paid engagement or testimonial.
MANYBOT 5.11.0 · EXTERNAL CONFIRMATION
Restart and retry produced opposite delivery failures.
After a passing offline native baseline of 283/283 tests, controlled lifecycle probes found two related behaviors around duplicate handling.
The maintainer explicitly confirmed this was a real issue rather than intentional behavior and requested the source locations, probe, repair sequence and acceptance tests. The maintainer also stated the open-source project was not seeking a commercial partnership or paid engagement; we preserve that distinction.
Source evidence: direct maintainer correspondence retained internally. This is external technical confirmation, not customer validation.
Evidence levels are intentionally different.
Behavior appeared under a controlled test.
The behavior survived a repeatable case.
An independent counterparty reproduced or confirmed it.
A customer paid for the outcome. We have not claimed this yet.