# Simple Reminders Standalone Meaning Audit - 2026-09-11 ## Scope and test This audit began with the 113 production reminders in thematic folders. It excludes the pre-existing `reminders/unverified/` corpus, the agent README, `reminders/welcome.md`, examples, and parser fixtures. During remediation, the Hans Moravec record was demoted from the thematic folders because its primary-source quotation boundary could not be established. The test is deliberately stricter than source authenticity: **does the quotation itself state a recognizable subject and proposition when encountered on a poster?** A source note or headline may clarify provenance, but it cannot substitute for a missing antecedent inside the poster-generating quotation. Unresolved `this`, `that`, `these`, `they`, or `it`; an unexplained percentage; a reaction without its target; and a clause cut away from the sentence that governs it are failures. ## Result - Original thematic reminders examined: **113** - Current verified reminders after provenance and semantic demotions: **66** - Current verified quotations that fail the standalone-meaning test: **0** - Records quarantined for standalone-meaning failure: **44** - Additional records quarantined for quotation-boundary or attribution failure: **3** - Original 113-item cohort now carrying `## Sources and Context`: **113 of 113** Passing this test does **not** establish verbatim accuracy, correct attribution, or adequate provenance. Those are separate gates. | Category | Quarantined for standalone failure | Remaining verified | |---|---:|---:| | AI Control | 17 | 15 | | Connectomics | 2 | 12 | | Consciousness | 2 | 8 | | Cybernetics | 0 | 1 | | Digital Afterlife | 2 | 1 | | Existential Risk | 9 | 7 | | Identity | 1 | 2 | | Machine Succession | 1 | 7 | | Mind Upload | 2 | 3 | | Preservation | 2 | 3 | | Risk Debate | 5 | 4 | | Simulation | 1 | 3 | | **Total** | **44** | **66** | ## Quotations quarantined for standalone failure ### AI Control - [[reminders/unverified/A Remarkable Understatement by Max Tegmark|A Remarkable Understatement by Max Tegmark]] — `it` does not identify the thing said to exceed a national-security threat. - [[reminders/unverified/AI Has Entered a Time of Huge Uncertainty by Geoffrey Hinton|AI Has Entered a Time of Huge Uncertainty by Geoffrey Hinton]] — the uncertainty and anticipated event are unnamed. - [[reminders/unverified/AI Is Accelerating Faster Than Expected by Yoshua Bengio|AI Is Accelerating Faster Than Expected by Yoshua Bengio]] — clipped causal fragment; the accelerating process is absent. - [[reminders/unverified/Can Humanity Stay in Charge of Advanced AI by Demis Hassabis|Can Humanity Stay in Charge of Advanced AI by Demis Hassabis]] — `those systems` has no antecedent. - [[reminders/unverified/Digital Intelligence Will Take Over by Geoffrey Hinton|Digital Intelligence Will Take Over by Geoffrey Hinton]] — `they` has no antecedent in the quotation. - [[reminders/unverified/Keeping Control of AI by Demis Hassabis|Keeping Control of AI by Demis Hassabis]] — `the systems` are not identified. - [[reminders/unverified/Let’s Slow Down by Yoshua Bengio|Let’s Slow Down by Yoshua Bengio]] — the object and reason for slowing down are absent. - [[reminders/unverified/More Intelligent Than Ourselves by Geoffrey Hinton|More Intelligent Than Ourselves by Geoffrey Hinton]] — `things` and `them` stand in for the advanced AI systems under discussion. - [[reminders/unverified/Nobody Knows What Advanced AI Will Do Next by Geoffrey Hinton|Nobody Knows What Advanced AI Will Do Next by Geoffrey Hinton]] — the quotation does not identify what is changing or what might happen. - [[reminders/unverified/Power Beyond Previous Technology by Sam Altman, Greg Brockman and Ilya Sutskever|Power Beyond Previous Technology by Sam Altman, Greg Brockman and Ilya Sutskever]] — `systems we are concerned about` omits the subject supplied by the source. - [[reminders/unverified/Proceed with Exceptional Care by Demis Hassabis|Proceed with Exceptional Care by Demis Hassabis]] — the thing that must proceed carefully is absent. - [[reminders/unverified/Racing Toward Self-Improving Superintelligence by Jacob Coxon|Racing Toward Self-Improving Superintelligence by Jacob Coxon]] — `they` does not identify the companies being criticized. - [[reminders/unverified/Stopping Superintelligence Could Require Global Surveillance by Sam Altman, Greg Brockman and Ilya Sutskever|Stopping Superintelligence Could Require Global Surveillance by Sam Altman, Greg Brockman and Ilya Sutskever]] — `it` lacks its source antecedent. - [[reminders/unverified/The Precautionary Principle for Advanced AI by Demis Hassabis|The Precautionary Principle for Advanced AI by Demis Hassabis]] — `the only right way to proceed` omits what action or program is at issue. - [[reminders/unverified/Think Hard About AI Before It Is Too Late by Christof Koch|Think Hard About AI Before It Is Too Late by Christof Koch]] — `this` has no antecedent. - [[reminders/unverified/Unexpected AI Acceleration Changed My Mind by Yoshua Bengio|Unexpected AI Acceleration Changed My Mind by Yoshua Bengio]] — the acceleration, letter, and `these topics` depend on omitted context. - [[reminders/unverified/When Machines Get What They Want by Stuart Russell|When Machines Get What They Want by Stuart Russell]] — `then` omits the condition that produces the stated outcome. ### Connectomics - [[reminders/unverified/Brain Circuitry and Mind by Edward S. Boyden|Brain Circuitry and Mind by Edward S. Boyden]] — terminates before completing the proposition. - [[reminders/unverified/The Brain-Mapping Task Humans Cannot Do Alone by Jan Funke|The Brain-Mapping Task Humans Cannot Do Alone by Jan Funke]] — `And that` omits the task Funke called impossible. ### Consciousness - [[reminders/unverified/If You Can Have Nonorganic Consciousness by Max Tegmark|If You Can Have Nonorganic Consciousness by Max Tegmark]] — conditional fragment without a consequence. - [[reminders/unverified/Large Language Models Already Understand by Geoffrey Hinton|Large Language Models Already Understand by Geoffrey Hinton]] — `These things` omits large language models from the quotation. ### Digital Afterlife - [[reminders/unverified/A Thousand-Year Journey by Randal A. Koene|A Thousand-Year Journey by Randal A. Koene]] — `It` does not identify a substrate-independent mind. - [[reminders/unverified/Mind Uploading Will Not Capture an Individual Mind by Olaf Sporns|Mind Uploading Will Not Capture an Individual Mind by Olaf Sporns]] — `This` does not identify the proposition Sporns rejected. ### Existential Risk - [[reminders/unverified/A 25 Percent Chance Things Go Very Badly by Dario Amodei|A 25 Percent Chance Things Go Very Badly by Dario Amodei]] — the percentage is authentic, but the event is not in the quotation; the source defines it as AI destroying humanity. - [[reminders/unverified/A Ten Percent AI Extinction Risk Is Not Unreasonable by Geoffrey Hinton|A Ten Percent AI Extinction Risk Is Not Unreasonable by Geoffrey Hinton]] — `it` does not identify AI-caused human extinction. - [[reminders/unverified/AI Could Be Humanity’s Best—or Worst—Event by Max Tegmark|AI Could Be Humanity’s Best—or Worst—Event by Max Tegmark]] — `it` omits AI. - [[reminders/unverified/Catastrophic AI Risk Deserves Serious Mitigation by Demis Hassabis|Catastrophic AI Risk Deserves Serious Mitigation by Demis Hassabis]] — terminal fragment without the risk being evaluated. - [[reminders/unverified/The AI Industry Is Not Reducing Risk Enough by Paul Christiano|The AI Industry Is Not Reducing Risk Enough by Paul Christiano]] — `this risk` is not defined in the quotation. - [[reminders/unverified/The Potential to Cause Immeasurable Harm by Max Tegmark|The Potential to Cause Immeasurable Harm by Max Tegmark]] — fragment without the system possessing the potential. - [[reminders/unverified/These Techniques Will Not Work for Superintelligence by OpenAI alignment researchers|These Techniques Will Not Work for Superintelligence by OpenAI alignment researchers]] — `these techniques` are not identified. - [[reminders/unverified/Turing Was Right to Worry—but Doom Is Not Inevitable by Stuart Russell|Turing Was Right to Worry—but Doom Is Not Inevitable by Stuart Russell]] — Turing's concern is not named. - [[reminders/unverified/Up to and Including Human Extinction by Stuart Russell|Up to and Including Human Extinction by Stuart Russell]] — terminal fragment without risk, actor, or causal claim. ### Identity - [[reminders/unverified/A Continuous Stream of Consciousness by David J. Chalmers|A Continuous Stream of Consciousness by David J. Chalmers]] — `the process` does not identify gradual uploading. ### Machine Succession - [[reminders/unverified/Hinton’s Guess — the Machines Will Take Over by Geoffrey Hinton|Hinton’s Guess — the Machines Will Take Over by Geoffrey Hinton]] — `they` lacks its antecedent. ### Mind Upload - [[reminders/unverified/A Complete Brain Simulation Would Probably Produce a Mind by Anders Sandberg|A Complete Brain Simulation Would Probably Produce a Mind by Anders Sandberg]] — both `it` and `that` refer to a question omitted from the quotation. - [[reminders/unverified/Substrate-Independent Minds by Randal A. Koene|Substrate-Independent Minds by Randal A. Koene]] — `This new substrate` is not identified. ### Preservation - [[reminders/unverified/Brain Preservation Could Carry People Beyond Death by Kenneth Hayworth|Brain Preservation Could Carry People Beyond Death by Kenneth Hayworth]] — `such a procedure` depends on an omitted description of long-term brain preservation. - [[reminders/unverified/Put Them on Pause Until We Can by Ariel Zeleznikow-Johnston|Put Them on Pause Until We Can by Ariel Zeleznikow-Johnston]] — neither `them` nor the treatment postponed by `until we can` is present. ### Risk Debate - [[reminders/unverified/AI Doom Scenarios Are Complete B.S by Yann LeCun|AI Doom Scenarios Are Complete B.S by Yann LeCun]] — reaction without the claim being rejected. - [[reminders/unverified/An Instant AI World Takeover Is Preposterous by Yann LeCun|An Instant AI World Takeover Is Preposterous by Yann LeCun]] — `the preposterous scenario` is absent from the quotation. - [[reminders/unverified/Debating AI Risk by Yann LeCun|Debating AI Risk by Yann LeCun]] — `it` does not identify the system whose safety is being evaluated. - [[reminders/unverified/Extinction I Don’t Know by Gary Marcus|Extinction I Don’t Know by Gary Marcus]] — clipped answer without the question or a stated probability judgment. - [[reminders/unverified/Why AI Extinction Risk Still Looks Unconvincing by Andrew Ng|Why AI Extinction Risk Still Looks Unconvincing by Andrew Ng]] — `it` does not identify AI. ### Simulation - [[reminders/unverified/A Complete Brain Simulation Could Do Everything by Henry Markram|A Complete Brain Simulation Could Do Everything by Henry Markram]] — `everything` does not identify the abilities attributed to a complete simulated brain. ## Comparison with the supplied provenance fragments The three supplied fragments are not mutually consistent. The broad 113-item wording list says it supplies a recoverable form for every item, while the later provenance audit explicitly retracts that level of confidence and identifies only **52 complete quotation units it can presently defend**. The later audit is the controlling comparison because it distinguishes verbatim speech from article titles, institutional descriptions, editorial summaries, and journalistic narration. Under that later audit: - **52** complete quotation units are presently defensible. - **61** of the original 113 production records are not independently supported there as complete, correctly bounded quotations. - **44** records failed the separate standalone-meaning test and were moved to `reminders/unverified/` with the defect preserved in each record and in every existing wiki placement. - These groups overlap; they must not be added together. The audit also specifically rejects or reclassifies production records built from Hayworth's essay subtitle, a secondary summary of Sebastian Seung, journalistic narration about Geoffrey Hinton, the tail of Andrew Payne's roadmap sentence, and clipped institutional statements attributed personally to Joergen Kornfeld. It leaves the Hans Moravec wording provenance-pending. Those are quotation-boundary or attribution defects even where the words themselves sound meaningful. The separate discovery ledger covered the 387 records that were already in quarantine when the scan began: **184** produced an exact-text search candidate, **17** produced only a prefix candidate, and **186** produced no match. An exact search hit is not verification; many hits are quotation aggregators, copied pages, or sources that repeat a saying without establishing its first publication. These records remain in `reminders/unverified/` until their original source, attribution, date, and complete thought boundary are resolved. ## Immediate corrections completed - Replaced Jan Funke's contextless reaction with his complete statement about why useful predictions were not enough without confidence that the network was actually working; synchronized [[wiki/Jan Funke|Jan Funke]]. - Restored Andrew Payne's complete 2024 mammalian-connectomics sentence; synchronized [[wiki/Whole Mouse Brain Connectome|Whole Mouse Brain Connectome]]. - Restored the complete four-author 2025 E11 Bio PRISM sentence; synchronized [[wiki/Johan Winnubst|Johan Winnubst]]. - Restored the preceding testing sentence to Dario Amodei's Responsible Scaling Policy quotation; synchronized [[wiki/Operational Autonomy|Operational Autonomy]]. - Corrected Amodei's Axios wording, date, venue, and probability context; synchronized [[wiki/Alignment Problem|Alignment Problem]]. The quotation remains in the review queue because its poster text says `things go really, really badly` rather than naming AI-caused human extinction. - Rejoined Evan Hubinger's extinction event and probability into one contiguous statement; synchronized its existing wiki placements. - Restored the complete Susan Schneider, Christof Koch, Vernor Vinge, Alan Turing, Stephen Hawking, OpenAI, Andrew Ng, Gary Marcus, Henry Markram, and Kenneth Hayworth passages supplied by the provenance audit, and propagated each changed quotation to its existing wiki rendering. - Replaced the non-quotation TED editorial summary attributed to Sebastian Seung with his spoken statement that experience changes the connectome; synchronized [[wiki/Sebastian Seung|Sebastian Seung]]. - Corrected `Your brain mediates everything that you sense, feel, think, and do` from a personal Edward Boyden attribution to the MIT Synthetic Neurobiology Group's institutional research description; synchronized [[wiki/Whole-Brain Connectomics|Whole-Brain Connectomics]]. - Restored the full Kornfeld Lab and Max Planck institutional statements, corrected their attribution away from personal speech, and synchronized their three existing wiki placements. - Demoted the Hans Moravec record to `reminders/unverified/` because the secondary source does not establish that the wording is Moravec's own sentence; preserved that qualification in [[wiki/Distributed Intention|Distributed Intention]]. - Demoted Hayworth's essay subtitle and the University of Toronto's narrated “passing stage” sentence because neither is a clean direct quotation; preserved their attribution boundaries in [[wiki/Brain Preservation Foundation|Brain Preservation Foundation]] and [[wiki/Civilizational Payload|Civilizational Payload]]. - Moved all 44 remaining standalone-meaning failures to `reminders/unverified/`, added the specific defect to each record, and propagated the same qualification and new path to every existing wiki placement. ## Editorial disposition Do not promote any remaining item merely by attaching a source URL. Each entry must pass three independent gates: verbatim provenance, a defensible quotation boundary, and standalone semantic value. Where the source contains no worthwhile self-contained quotation, retire or replace the reminder rather than allowing its headline or source note to manufacture meaning the quotation does not contain.