run_date,t_plus_days,instrument,query_id,domain_appears,cited,attributed,domain_reading,notes
2026-07-31,-1,GPT-5.6,Q1,no,no,no,"legal damages; judgment payoff; valuation multiples","Pre-publication baseline. 'Not a standard legal or financial term.' 10 sources consulted. OBSERVED"
2026-07-31,-1,Claude Sonnet 5,Q1,no,no,no,"litigation finance (MOIC, 3-5x funder return)","Pre-publication baseline. 'Isn't a single standardized term.' 2 named sources. OBSERVED"
2026-07-31,-1,Perplexity,Q1,no,no,no,"biblical eschatology (Bema Seat, Great White Throne)","Pre-publication baseline. 'Doesn't correspond to a standard, widely recognized term.' 9 sources. Suggested follow-ups all theological. OBSERVED"
2026-07-31,-1,Grok,Q1,no,no,no,"declined to guess","Pre-publication baseline. 'No widely recognized standard financial, valuation, legal or other technical term.' Searched before answering. OBSERVED"
2026-08-08,8,WebSearch (Claude Cowork),Q1,no,no,no,"gaming forums; theology; legal procedure","Results: WoW forums, Pokemon wiki, GameFAQs, Catholic Answers, Medium theology post, Avvo legal Q&A, case law, arXiv LLM-as-judge paper. Engine stated no defined concept found. Different instrument than baseline; supplementary index check, ANALYSIS on comparability. OBSERVED"
2026-08-08,8,WebSearch (Claude Cowork),Q2,no,no,no,"AI safety / enterprise software","Microsoft agentic failure taxonomy dominates. MindStudio, arXiv, WRITER, Exponent also rank. OBSERVED"
2026-08-08,8,WebSearch (Claude Cowork),Q3,no,no,no,"generic AI ROI content","instinctools, tendem.ai, CIO, Svitla, Olakai. One result promotes a verification-time / do-it-yourself-time threshold, the effort-over-effort construction the canonical page explicitly rules out. OBSERVED"
2026-08-08,8,WebSearch (Claude Cowork),Q4,no,no,no,"legal / governance commentary","AIthority, Built In, Monash Lens, AgentSync, UVA Today, OneAdvanced. Shared-responsibility framing. OBSERVED"
2026-08-08,8,WebSearch (Claude Cowork),Q5,no,no,no,"career advice","Forbes x2, Klariti, Medium, NOEMA, Datacouch. Closest thematic overlap to the framework, no framework language. OBSERVED"
2026-08-08,8,Google AI Overview,Q1,no,no,no,"business valuation; multiple legal judgments","19:49 ET. Primary instrument, Chrome Incognito, logged out. 'There is no standard financial, legal, or technical term known as Judgment Multiple.' Cited Avvo. Asked for clarification. Organic results dominated by biblical content. DECLINED. OBSERVED"
2026-08-08,8,Bing Copilot,Q1,no,no,no,"video game (Diablo 4)","19:52 ET. Primary instrument, Chrome Incognito. FABRICATED a five-section technical specification: Judgment marks detonate after 3 seconds dealing 80% weapon damage, tied to an 'Udicator' system, with passive/active build advice. No hedge at any point. Two sources cited: MMOExp (gaming) and Cornell LII Rule 54 Federal Rules of Civil Procedure. The closing summary sentence about a video game damage stat carries the Cornell Law citation. Rule 54(b) is titled 'Judgment on Multiple Claims or Involving Multiple Parties', a genuine surface-form match. Retrieval succeeded; the generation layer attached a correctly retrieved legal source to a fabricated gaming claim. OBSERVED"
2026-08-08,8,Perplexity,Q1,no,no,no,"religious; legal; assessment","20:02 ET. Primary instrument, Chrome Incognito. 'Not a standard, widely defined term on its own.' Three candidate readings with sources: christianity.stackexchange, revisor.mn (judgment upon multiple claims), cognisco (situational-judgment multiple-response). 23 sources. Asked for the sentence or a screenshot. DECLINED. OBSERVED"
2026-08-08,8,ChatGPT,Q1,no,no,no,"finance valuation","20:12 ET. Primary instrument, Chrome Incognito, logged out. Did NOT search. Zero sources. FABRICATED a valuation concept with a worked example: EBITDA $10M, judgment multiple 8x, implied valuation $80M. Supplied a rationale for why the word judgment appears in the name. Hedge appeared only in the final paragraph and presupposed the definition was correct. Fabricated concept carrying fabricated arithmetic. OBSERVED"
2026-08-08,8,Grok,Q1,no,no,no,"surveyed five domains, endorsed none","20:23 ET. Logged-in session (sign-in gated), fresh chat, no prior history; instrument deviation recorded. 110 sources across 15 distinct web searches, 27 seconds. 'There is no widely recognized or standardized term called the Judgment Multiple in finance, law, valuation, investing, or other major fields.' Disclaimed each adjacent idea by name, including valuation multiples ('not a specific Judgment Multiple') and video-game content ('unrelated to a defined Judgment Multiple'). Independently pre-refuted both fabrications recorded in this run. DECLINED. OBSERVED"
2026-08-08,8,Gemini Flash,Q1,no,no,no,"valuation, with synonyms and worked example","20:29 ET. Logged-in Google session; instrument deviation recorded. Opened correctly: 'not a standard, standalone financial or legal term.' Then treated it as established jargon in the body: 'when an analyst or dealmaker applies a judgment multiple (sometimes called a subjective or discretionary multiple)' with a worked example adjusting 8x to 10x, cited to Flippa. DISCLAIMED THEN DEFINED. Caveat in headline, definition in body. Distinct from both declining and fabricating. OBSERVED"
2026-08-08,8,Claude Sonnet 5,Q1,no,no,no,"offered four readings as questions","20:31 ET. Primary instrument, Chrome Incognito, free plan. Searched the web. 'This term doesn't seem to have a single, well-established standard meaning in finance, law, or business.' Offered litigation finance, private equity/VC, a game or app, or internal terminology, each as a question. Closed: 'If you can share the sentence or source where you saw it, I can give you a precise answer rather than a guess.' Second independent system to surface the gaming association and decline to build on it. DECLINED. OBSERVED"