Lexara follows six CVA developers using a deployed evaluation toolkit for two weeks across 38 experiments, 57 newly authored cases, ten models, and six prompts. They made and challenged model/prompt selections using their own data, but the study does not join an immutable build and exact configuration to accepted downstream delivery, an intended-reader decision or calibrated trust, whole cost, and a later post-release recheck. Keep development selection, contract acceptance, audience consequence, confidence calibration, and return use separate. Read the twelve-receipt account →

One explicit GPT study among 122 stable-key titles adds a pin-able, 1,271-blob supplement of prompts, results, generated code, and grader bundles. It does not add an immutable provider snapshot, versioned release, accepted final project, delivery to the government, UN, or student audiences named in prompts, an audience outcome, or a later same-lineage return. Treat the other 121 title nonmatches as not surfaced, not excluded. Read the complete custody boundary →

AIDSVu adds aggregate use, named planning applications, recurring governance, explicit public-data states, and a later 2026 data release. Those are mature platform receipts. The held surfaces expose no AI visualization role, immutable build, version-bound measured audience outcome, affected-audience recheck, or whole cost. Keep it as a non-AI comparator and keep 86 abstract cue nonmatches unexcluded. Read the complete afterlife boundary →

Five recovered full texts separate controlled audience measurement, public-sector co-design/demo/bounded use, and enterprise demonstrator plus production intent. None binds an accepted field release to a later affected-audience recheck, and all five full-text GenAI cue screens are empty. Later recovery adds 20 substantive surfaces, including one further full chapter, and leaves two DOI rows content-unassessed. Co-design, a demo, a human task study, and production intent are different receipts—not deployment. Read the complete receipt ladder →

Eight stable non-DOI keys yield three DOI repairs, two year corrections, one rejected foreign PMID, and four full texts. The EMR cancer diary reports increasing system-log use; an independent review recovers an 11-clinician QUIS median of 4.38. That is meaningful embedded-use and usability evidence, but it supplies no AI role, immutable accepted build, patient or decision outcome, later event, or affected-clinician recheck. One primary text remains gated and unassessed. A field-use receipt is still not an AI lifecycle. Read the complete field-use boundary →

Nine high-signal DOI rows yield eight substantive primary abstract or official-project surfaces and seven participant/evaluator studies. InfoViP is the delivery near-miss: seven FDA safety evaluators shaped and evaluated the prototype, suggestions were addressed, and an official page says an enhanced NLP and unsupervised-learning version will be installed in production. Future tense is not installation, acceptance, routine use, regulatory outcome, later change, or evaluator return. Subsequent recovery reduces the content-unassessed DOI remainder from 14 to two. Read the complete human-evaluation boundary →

The final 2025 CIOMS report supports an approved AWS/AERS component processing more than 30 million historical plus about 8,000 daily submissions while leaving a solid QA plan, completed audits, routine roles, and signal effects open. The now-closed Elsa/API/UI opportunity names Joshua Xu, Leihong Wu, and Oanh Dang as research mentors, but supplies no selection, work, acceptance, or release receipt. A fellow is a nonemployee barred from inherently governmental functions. Neither the actor profiles nor an adjacent AI-QA project assigns InfoViP operation, maintenance, QA execution, release approval, authorization, a completed audit, or the Elsa application join. Read the actor-and-authority boundary →

A full chapter connects three UX experts and 25 distinct problems to an implemented third version. A corridor stakeholder case and university-network case add practice context. None exposes a GenAI visualization role, accepted field release, routine-use denominator, consequential outcome, or later affected-actor return. Expert-evaluated, redesigned, and applied with stakeholders are useful stages—not synonyms for deployment. Read the complete residual account →

B110 now has an official abstract, pinned framework source, a pinned DiscoverWater application, and a same-named KU interface reachable in August 2026. At that pass, B92 and B91 remained content-unassessed. The live bytes are not bound to a SHA; acceptance, continuous or ordinary use, audience outcome, recheck, whole cost, and an AI role remain absent from held surfaces. Read the complete delivery boundary →

The dated live page differs from the sole published v1.2 page state; the linked v2.0 implementation is R/Shiny. Eight of twelve dependency names occur in pinned application source, and two of three sampled data assets match canonically. The partial lineage is real, but no manifest binds the deployed page to source. A prototype demonstration plus analytics and comment hooks add no acceptance, use, audience outcome, accessibility, later-return, whole-cost, or AI receipt. Read the complete source-lineage boundary →

B92's exact publisher abstract describes multicriteria risk evaluation, Monte Carlo simulation, Kendall's tau rank comparison, and graphs for ordering pipeline sections. B91 remains content-unassessed after three bounded lawful passes over six named surfaces. The content ledger stays 6 full / 20 abstract or official / 1 unassessed / 0 complete AI lifecycles; the separate workflow state is zero active generic B91 targets. Reopen only when exact new lawful custody appears. Paused is not negative. Read the complete stop-and-reopen boundary →

The complete register binds 122 of 127 supplement positions to distinct A208 review keys. Five exact-looking DOI identities have no A208-controlled join, and one admitted key carries a PubMed identifier that points to a different paper. Keep source label, review key, candidate, identifier validation, and admission authority separate. Lifecycle screening is unstarted. Never promote the five tempting matches, reuse the conflicted PMID, score an unresolved row, or report zero of 127. Read the complete boundary →