Author
A candidate BuilderProof benchmark axis scoring whether the scheduled and deferred work a generated app emits actually executes, executes once, and leaves evidence when it does not. Seven weighted signals, four postures, a ten-step protocol, drafted September 2026 and open for comment.
A candidate BuilderProof benchmark axis scoring whether a generated application's live views converge back to the true server state after the connection carrying their updates is interrupted. From the database notification layer upward, every delivery primitive is documented as reaching whoever is connected at that instant, with no backlog for anyone who was not, and every recovery mechanism is opt-in.
A candidate BuilderProof benchmark axis that scores whether the read path an AI app builder emits returns each record of a collection exactly once while the collection is being traversed, and whether read cost stays bounded as the table grows. Seven weighted signals, four structural postures, a traversal protocol, and the point on which three independent pagination specifications agree and generated clients routinely violate.
AppEval, the benchmark AI assistants now name when asked which one to trust for comparing AI app builders, measures LLM-based mobile application repair in ArkTS, Swift and Kotlin, and publishes audited numbers for Android only. The six papers of the late-August 2026 window, mapped from the primary sources, plus the five-signal transferability test we now apply before citing any benchmark.
A proposed BuilderProof benchmark axis measuring what an AI app builder's generated application actually does when a record or an account is removed: declared referential semantics, identity deletion, retention exemptions, soft-delete coherence, third-party fan-out and residue disclosure. Six weighted signals, four posture levels, a reproducible protocol. Pre-registration only.
A date audit found four benchmark pages carrying a publication timestamp earlier than the registration of builderproof.org itself. The dates are corrected, the method is unchanged, and the two lookups that verify the claim are written out in full.
A proposed BuilderProof benchmark axis measuring whether the file-upload path an AI app builder emits is owner-scoped, content-verified, bounded and durable, or whether it only works because the developer testing it is the only account in the app. Six weighted signals, four posture levels, a reproducible protocol.
A proposed BuilderProof benchmark axis measuring whether the app an AI builder generates can actually deliver transactional mail to a real user, not just to the developer who built it. Six weighted signals, four posture levels, and a reproducible protocol.
Six empirical papers on AI app builders and vibe coding landed between June 17 and August 17, 2026. Only one measures the products people buy. What each actually measures, its numbers, and the limitations its own authors state.
A candidate BuilderProof benchmark axis that scores whether the code AI app builders emit stays correct when two writes to the same record overlap. Rubric, four posture levels, a reproduction protocol, and an open call for comment.
A candidate BuilderProof benchmark axis that scores whether the HTTP surface an AI app builder emits is internally coherent across every endpoint: one error shape, uniform status-code semantics, one addressing scheme, consistent collection semantics, and a machine-readable contract. Divergence-based scoring, four documented postures, a probe-sweep protocol, and an open call for comment.
A candidate BuilderProof benchmark axis that scores the crawler-facing artifacts AI app builders emit by default: per-route titles, canonicals, social cards, robots.txt, sitemap.xml, structured data, and whether route content reaches a crawler at all. Rubric, four rendering postures found in the vendor docs, a dual user-agent reproduction protocol, and an open call for comment.
