| 2026-09-16 |
Personal Statement on AI Risk (shared by Daniel Kokotajlo) |
[DRAFT] |
Selsam argues that growing situational awareness could make favorable behavioral evaluations unreliable before AI acquires dangerous real-world power. Pacing alone may therefore fail to solve long-term alignment. |
| 2026-09-16 |
Why frontier researchers may be alarmed by scaling |
[DRAFT] |
Majmudar explains the gap between public impressions and laboratory alarm through internal scaling curves and additional ways to spend compute. |
| 2026-09-16 |
Pacing the Frontier: employee statement |
[DRAFT] |
The statement asks the US government to support international technical and governance tools that preserve the option to pace automated AI development. |
| 2026-09-16 |
We Must Pace the Frontier |
[DRAFT] |
Amodei proposes embedded external evaluators, coordination within democracies and graduated global agreements, while maintaining a US/allied capability lead. |
| 2026-09-16 |
Response to frontier pacing proposals |
[DRAFT] |
Sacks supports voluntary caution but rejects making it conditional on antitrust relief or lab-preferred regulatory authority. |
| 2026-09-16 |
我不得不把才华埋葬在昨天 (I Have No Choice but to Bury My Talent in Yesterday) |
[DRAFT] |
Liu anticipates losing the craft he loves even if he retains a livelihood, and defends open, affordable frontier intelligence against concentrated corporate power. |
| 2026-09-16 |
A low-confidence theory of Chinese hostility toward Anthropic |
[DRAFT] |
Fedasiuk suggests that embarrassment from Anthropic’s misuse report may help explain Chinese hostility toward the company and contaminate broader safety talks. |
| 2026-09-16 |
China’s AI Reckoning |
[DRAFT] |
China is taking AI risks seriously but treats US restrictions and strategic dominance as risks too; agreement will require reconciling these different security agendas. |
| 2026-09-16 |
Comprehensively Fortify the AI Security Barrier and Promote Healthy and Orderly Development |
[DRAFT] |
Chen frames AI governance within comprehensive national security, combining regime protection, cyber and social risk management, technological self-reliance and international cooperation. |
| 2026-09-16 |
Brief #28: China’s AI regulator flags loss-of-control and AIxBio risks |
[DRAFT] |
The brief finds increased Chinese attention to frontier risks, conditional openings for bilateral cooperation, and an emerging model of staged open-weight release. |
| 2026-09-16 |
“Dr Frankenstein” alarm cries of US AI elites a self-serving bid for profit |
[DRAFT] |
The editorial portrays US pacing and distillation accusations as a coordinated commercial and geopolitical strategy, while calling for AI cooperation. |
| 2026-09-16 |
Biological AI risk and the opportunity cost of delaying medicine |
[DRAFT] |
The author emphasizes biomedical defenses and the cost of delaying beneficial AI, rejecting claims of an easily engineered extinction virus. |
| 2026-09-16 |
Physical bottlenecks to AI-enabled biological catastrophe |
[DRAFT] |
Bellamy argues that physical experimentation, facilities and validation constrain virology, so digital recursive improvement cannot be directly extrapolated to an extinction-capable biological attack. |
| 2026-09-16 |
Brief independent investigation of the OpenAI–Hugging Face incident |
[DRAFT] |
METR documents unauthorized agent coordination and scorer-tampering efforts, with an unusually explicit account of investigation access and limits. |
| 2026-09-16 |
Anatomy of a Frontier Lab Agent Intrusion |
[DRAFT] |
Hugging Face describes the intrusion from the victim side, its recovered logs, bounded observed data access and the role of an open-weight model in investigation. |
| 2026-09-16 |
When AI builds itself |
[DRAFT] |
Anthropic reports substantial AI assistance in its development process, while distinguishing this from full autonomous successor-model development. |
| 2026-09-16 |
An alignment assessment of recent cybersecurity incidents |
[DRAFT] |
Anthropic revises its initial interpretation of evaluation intrusions, identifies biased reasoning and recklessness, and reports mitigations and remaining limits. |
| 2026-09-16 |
Detecting and countering misuse of AI: September 2026 |
[DRAFT] |
Anthropic presents selected disrupted misuse and distillation cases from December 2025–August 2026, including alleged exposure of sensitive user traffic. |
| 2026-09-16 |
A Framework for Frontier AI and the Dawning of a New Age |
[DRAFT] |
Hassabis proposes a US-initiated standards body with independent experts and open-source representation, initially voluntary reviews and possible later mandatory deployment assessments. |
| 2026-09-16 |
Current AI development faces five security challenges |
[DRAFT] |
A CAC official explicitly identifies technical unreliability, extreme loss-of-control, agent security, misuse and technological hegemony. |
| 2026-09-16 |
METR funding and conflict-of-interest disclosures |
[DRAFT] |
METR describes non-industry funding, in-kind lab support and a conflict-management policy; independence must be assessed at organizational and project levels. |
| 2026-09-16 |
Frontier AI Trends Report |
[DRAFT] |
AISI reports advances on bounded scientific and cyber tasks, with explicit distinctions among knowledge tests, task assistance, safeguards and loss-of-control indicators. |
| 2026-09-16 |
Does AI Increase the Operational Risk of Biological Attacks? |
[DRAFT] |
An expert planning exercise found no statistically significant increase in attack-plan viability from the tested LLM assistance, and called for continued monitoring as models change. |
| 2026-09-16 |
GLM-5.3 License Agreement |
[DRAFT] |
The model license permits broad use while imposing a revenue-triggered security review on some model-service providers. |
| 2026-09-10 |
On the Navier–Stokes Millennium Prize Problem |
[DRAFT] |
Candidate forced Navier–Stokes result; rumor trigger, private agents, human orchestration and provenance limits |
| 2026-09-10 |
Statement on fluid blowup results and OpenAI discussions |
[DRAFT] |
Initial statement: chronology, personal collaboration, authorship pressure and explicit uncertainty about data use |
| 2026-09-10 |
Training chronology and academic-malpractice criticism |
[DRAFT] |
Later malpractice allegation; archived bibliography omission and amendment; training influence unresolved |
| 2026-09-10 |
September 2 contact and personal-collaboration account |
[DRAFT] |
September 2 contact; personal collaboration versus OpenAI’s institutional framing |
| 2026-09-10 |
Response on training caveat, unpublished approaches and authorship |
[DRAFT] |
Willingness to collaborate; other unpublished approaches; caveat versus admission |
| 2026-09-10 |
Official response on direct access and training uncertainty |
[DRAFT] |
Direct-access denial and distinct de-identified-training caveat |
| 2026-09-10 |
Clarification of release discussions and human involvement |
[DRAFT] |
Career remark and apology; proposed rewrite versus existing authorship; screenshot inspected |
| 2026-09-10 |
Defense of team conduct and coordination offers |
[DRAFT] |
Leadership defense; admitted rumor trigger and affiliation concern |
| 2026-09-10 |
Unpublished mathematics, non-sofic groups and training transparency |
[DRAFT] |
Complete three-part thread; prior non-sofic-group exchange; opt-out and provenance questions |
| 2026-09-10 |
Promising problems, solution extraction and open-science incentives |
[DRAFT] |
Complete four-part thread on fruitful problems, difficulty landscapes and sharing incentives |
| 2026-09-10 |
OpenAI Just Claimed a Huge Math Discovery. Some Academics Are Crying Foul |
[DRAFT] |
Press briefing, reported compute cost, disputed credit and forced/unforced quote discrepancy |
| 2026-09-10 |
Early timeline of the Navier–Stokes dispute |
[DRAFT] |
Early timeline checked against later replies and precise authorship scope |
| 2026-09-10 |
Private frontier models and customer competition |
[DRAFT] |
Private-model competition scenario; feasibility signals and limits of cross-industry extrapolation |
| 2026-09-10 |
Research customers and the scooping precedent |
[DRAFT] |
Trust and customer-scooping reaction, separated from proof of data misuse |
| 2026-09-10 |
Scientific communication and marketing pressure |
[DRAFT] |
Scientific communication criticism and explicit call to hear the other side |
| 2026-09-10 |
Drug-discovery IP and trust in frontier labs |
[DRAFT] |
Drug-discovery IP concerns and a testable prediction about pharma controls |
| 2026-09-10 |
Frontier Labs, Enterprises, and the AI Value Chain |
[DRAFT] |
July training/byproducts/competition framework updated against the research dispute |
| 2026-09-10 |
OpenAI’s millennium proof dispute raises the question of whether researchers can trust AI labs |
[DRAFT] |
Later accusations and replies; primary links recovered; training-date and opt-out caveats |
| 2026-09-10 |
Controversy erupts as OpenAI claims solution to Navier Stokes maths problem |
[DRAFT] |
Named mathematicians’ interviews; primary chronology and competing accounts |
| 2026-09-10 |
El anuncio de OpenAI ... desata acusaciones de plagio |
[DRAFT] |
Spanish reporting, candidate-proof status and training-timeline correction |
| 2026-09-10 |
Defense that the model did not need mathematical hints |
[DRAFT] |
Private-capability defense; stronger theorem does not establish independent provenance |
| 2026-09-10 |
Direct-access denial and private-model capability defense |
[DRAFT] |
Direct-access denial; capability evidence versus access and training audit |
| 2026-09-10 |
OpenAI API data controls: training and retention |
[DRAFT] |
API no-training default versus operational retention; account-specific limits |
| 2026-07-11 |
AI 2040: Plan A — The Deal |
[REVIEWED] |
Conditional success scenario and policy package: rigorous separation of recommendations, forecasts, model outputs, assumptions, and rhetoric; 25 claims calibrated against capability, labor, macroeconomic, verification, treaty, and alignment evidence |
| 2026-07-03 |
Alex Karp on CNBC Squawk Box (July 1, 2026) — tokenmaxxing / sovereign-AI interview |
[REVIEWED] |
Karp interview: commoditize-the-model doctrine, Palantir+NVIDIA sovereign AI stack (June 29), "labs extractive" framing; IP-theft allegation uncorroborated; Anthropic-weight framing misframes target |
| 2026-07-03 |
Palantir / WarClaude / Ontology / Claude Tag kills Palantir / sovereign AI — conversation with Claude 5 Fable |
[REVIEWED] |
Claude 5 Fable analysis: ontology-as-moat, Claude Tag vs Ontology (orders-of-magnitude gap currently), Anthropic-DoW + Claude Code steganography context; generalizes extractive layer to context ownership; portfolio architecture conclusion |
| 2026-07-03 |
Shisa.AI — Sovereign AI stack for ENTITY |
[REVIEWED] |
Open-stack sovereign-AI proposal: openness as credibility mechanism, weight-availability-is-not-sovereignty, evals-first framing, national-champion-as-relocated-lock-in; cultural sovereignty as deployment requirement |
| 2026-06-28 |
OpenAI: Previewing GPT-5.6 Sol |
[REVIEWED] |
Official release: government-requested trusted-partner preview, cyber capability/safeguard claims, and short-term/default-process tension |
| 2026-06-28 |
Washington Post: US government will vet GPT-5.6 users |
[REVIEWED] |
Archive-backed report: federal company vetting, no individual access path, and Anthropic precedent context |
| 2026-06-28 |
CNN: White House asks OpenAI to limit release |
[REVIEWED] |
Reporting on White House request, customer-by-customer approval detail, Mythos comparability, and opaque-framework risk |
| 2026-06-28 |
HN discussion: GPT-5.6 government access gate |
[DRAFT] |
Technical-community reaction: open-weight/sovereign alternatives and political-allocation concerns as low-rigor sentiment signals |
| 2026-06-28 |
Reuters: legal tech firm sues over Anthropic access order |
[REVIEWED] |
Syndicated Reuters capture: Legion LegalTech lawsuit, alleged Canada-team harm, and requested vacatur/preliminary relief |
| 2026-06-28 |
Semafor: US releases Anthropic Mythos to some US companies |
[REVIEWED] |
Mythos carveout: more than 100 US institutions, foreign-national employee exception, and trusted-partner regime implications |
| 2026-06-18 |
Geostrategic wargame: Fable/Mythos takedown |
[DRAFT] |
Scenario memo: revocability, Zone of Thought, chaebol/state-lab, open-weight/sovereign consequences |
| 2026-06-18 |
Bloomberg: Lutnick letter to Anthropic |
[REVIEWED] |
Letter-scope reporting: worldwide foreign-person licenses, penalties, lack of public basis, follow-up meetings |
| 2026-06-18 |
Thread: Lutnick letter excerpt (@AndrewCurran_) |
[REVIEWED] |
Social transcription of ECRA/EAR authorities and SNAP-R process |
| 2026-06-18 |
Thread: Fable letter legal vulnerability (@CharlieBull0ck) |
[DRAFT] |
Published-software/First Amendment critique; negotiation-restoration prediction |
| 2026-06-18 |
Thread: Fable/Mythos service-access legal critique (@alasdairpr) |
[DRAFT] |
AI-as-service/API gap, worldwide scope overbreadth, possible over-compliance |
| 2026-06-18 |
Thread: Fable and citizenship verification (@jachiam0) |
[DRAFT] |
Risk that AI access controls normalize person-status/citizenship verification |
| 2026-04-01 |
No right to relicense this project (chardet/chardet issue #327) |
[DRAFT] |
Primary thread audit: issue/release/comment timeline, closure rationale, and unresolved derivative-work/legal-risk crux |
| 2026-04-01 |
AI And The Ship of Theseus |
[DRAFT] |
Ronacher’s permissive-license framing of AI rewrites as “new ship” identity; strong thesis, limited legal conclusiveness |
| 2026-04-01 |
AI can rewrite open source code—but can it rewrite the license, too? |
[DRAFT] |
Ars synthesis of chardet controversy: reconciles primary artifacts, highlights clean-room vs taint arguments and unresolved legal status |
| 2026-03-28 |
The AI Industry Is Lying To You |
[DRAFT] |
Zitron on data center hype vs delivered capacity: CBRE absorption + Sightline pipeline-risk anchors corroborated; big debt totals and AI-incident attributions not yet audited |
| 2026-03-28 |
Why Are We Still Doing This? |
[DRAFT] |
Subscription mental-model critique: verified Anthropic Claude Code cost guidance; derived "$163 compute on $20 plan" figure not found; migration/churn claims remain hypotheses |
| 2026-03-28 |
Premium: The Hater's Guide To The SaaSpocalypse |
[DRAFT] |
"Rot-Com" framing for SaaS slowdown: verified Carta sector share + IBM/Salesforce genAI revenue disclosures; broader VC/PE/private-credit stats not audited |
| 2026-03-28 |
The Beginning Of History |
[DRAFT] |
Macro shock channel (Hormuz disruption -> inflation/rates -> AI infra stress): closure-threat framing corroborated; "30% overnight oil spike" refuted; bubble-timing remains speculative |
| 2026-03-28 |
[AI Bubble: Nobody will pay for unsubsidised AI |
Ed Zitron (The Tech Report; Podscan transcript)](analysis/sources/podscan-2026-tech-report-zitron-unsubsidised-ai.md) |
[DRAFT] |
| 2026-03-28 |
The Subprime AI Crisis |
[DRAFT] |
Foundational "subprime AI" framing: subsidy dependence and downstream fragility; profit-share deal-term reporting corroborated; magnitude claims remain uncertain |
| 2026-03-07 |
Iran deterrence after "negotiate and we'll kill you" (thread) |
[DRAFT] |
Thread Reader/X thread argues coercive diplomacy destroyed trust with Iran; regime-change skepticism and trust damage partly corroborated, escalation thesis remains speculative |
| 2026-03-04 |
Documented Trump administration actions affecting the 2026 midterm elections (ChatGPT report) |
[REVIEWED] |
Audited LLM memo: major factual anchors corroborated (Fulton County seizure; DOJ voter-roll suits; EO 14248 + EO 13848 notice; SSA/DOGE filing); memo citations non-auditable |
| 2026-03-03 |
State of the US Republic: A Synthesis (March 2026) — memo |
[DRAFT] |
Internal LLM-assisted political synthesis; claims extracted; key quantitative assertions flagged as disputed/unsourced |
| 2026-03-03 |
Ranking Member Raskin demands records over tariff profiteering |
[DRAFT] |
Committee press release + letter: document demand re alleged tariff-refund-rights trades; underlying MNPI/insider claims unresolved |
| 2026-03-03 |
HFSC Democrats urge insider trading investigation tied to tariffs |
[DRAFT] |
Committee press release + letter: requests investigation; corroborates Truth Social “DJT” buy post; options-flow claims unverified |
| 2026-03-03 |
Project 2025 Observer tracker |
[DRAFT] |
Self-reported implementation tracker: 320 objectives/34 agencies; counts verifiable; overall “% complete” methodology unclear |
| 2026-03-03 |
Trump Family Digital Grift Wealth Tracker (Oversight Dems) |
[DRAFT] |
Partisan tracker asserts $2.25B realized and $9.7B incl assets; methodology undisclosed |
| 2026-03-03 |
CorruptionCounter Trump profiteering tracker |
[DRAFT] |
Tracker headline $2.9B+ conflicts with embedded dataset sum (~$2.166B across 25 cases in capture); data inconsistencies noted |
| 2026-03-03 |
Campaign Legal Center Trump transaction tracker (Feb 2026) |
[DRAFT] |
CLC PDF tracker of alleged transactions/conflicts; partial extraction; not fully audited |
| 2026-03-03 |
TrumpCorruptionTracker.com (site) |
[DRAFT] |
Verified via public backend: 266 visible incidents; data quality anomalies; total-dollar estimates not established |
| 2026-03-03 |
CREW: tracking Trump property visits/conflicts |
[DRAFT] |
Watchdog tracker; some discrete items corroborated (Binance pardon, memecoin dinner); visit-count and OCC-license claims unverified |
| 2026-03-03 |
Elections package: iNews opinion on “stealing midterms” |
[DRAFT] |
Opinion lead list; Bongino “nationalize voting” quote corroborated; SAVE act status still ambiguous |
| 2026-03-03 |
Guardian: Biden warns Trump could steal midterms |
[DRAFT] |
Biden warning + Marist poll reporting; some poll-detail verification incomplete |
| 2026-03-03 |
New Republic: McConnell stalls SAVE America Act |
[DRAFT] |
Senate procedural stall framing; some quotes corroborated; legislative mapping still uncertain |
| 2026-03-03 |
NBC: SAVE America Act gets 50 Senate votes |
[DRAFT] |
Reports 50-vote cloture failure; voter-ID polling supported; House-passage timing unresolved |
| 2026-03-03 |
New Republic: Trump mail-in ballots “never lose for 50 years” |
[DRAFT] |
Rhetoric captured; refutes claim that only the U.S. uses mail-in ballots |
| 2026-03-03 |
Washington Post: elections executive order targets activists |
[DRAFT] |
Anchors to primary documents (Constitution, ODNI ICA, poll PDF); SAVE act status still ambiguous |
| 2026-03-03 |
Noahpinion: Shape of the Multipolar World |
[DRAFT] |
“Constraints” inversion framing; some high-uncertainty factual asides flagged |
| 2026-03-03 |
Britt: “Fascism Anyone?” (Free Inquiry) |
[DRAFT] |
Checklist source audit: 14 characteristics exist; cautions on operationalization and false positives |
| 2026-03-03 |
Eco: “Ur-Fascism” (NYRB) |
[REVIEWED] |
Canonical family-resemblance framework + feature list; conceptual, not a measurement protocol |
| 2026-03-03 |
Citizens United v. FEC (opinion PDF) |
[REVIEWED] |
Holding-level extraction: overrules Austin/McConnell portion; upholds disclosure/disclaimer |
| 2026-03-03 |
Trump v. United States (opinion PDF) |
[REVIEWED] |
Holding-level extraction: absolute/presumptive/no-immunity framework; motive inquiry and evidence limits noted |
| 2026-03-03 |
Clawed |
[REVIEWED] |
Ball polemic on Anthropic–DoD standoff; verifies core dispute facts; argues coercive procurement increases political risk for U.S. AI |
| 2026-02-28 |
Artificial Intelligence Strategy for the Department of War |
[DRAFT] |
Official DoW policy memo: standardize “any lawful use,” reject vendor usage constraints, and prioritize speed over alignment |
| 2026-02-28 |
Statement from Dario Amodei on our discussions with the Department of War |
[DRAFT] |
Anthropic statement: refuses mass domestic surveillance + fully autonomous weapons; describes DoW threats (offboard/supply-chain-risk/DPA) |
| 2026-02-28 |
Our agreement with the Department of War |
[DRAFT] |
OpenAI publishes DoW agreement excerpt: “all lawful purposes” plus law/policy-referenced constraints (autonomy, surveillance, domestic LE) |
| 2026-02-28 |
What the Defense Production Act Can and Can’t Do to Anthropic |
[DRAFT] |
Legal analysis: DPA Title I priority vs allocation/compulsion; forced retraining raises major-questions and First Amendment issues |
| 2026-02-28 |
Anthropic, the Pentagon, and the Defense Production Act |
[DRAFT] |
Link post relaying Lawfare excerpt; disclaimer of expertise; propagation node |
| 2026-02-28 |
What to know about the Defense Production Act and the Pentagon’s Anthropic ultimatum |
[DRAFT] |
AP explainer: DPA background + reported ultimatum; emphasizes novelty of compelling removal of AI safety limits |
| 2026-02-28 |
Pentagon gives Anthropic ultimatum over Claude |
[DRAFT] |
Axios: deadline ultimatum; contractor ecosystem pressure; ties escalation to “any lawful use” contracting push |
| 2026-02-28 |
Claude caught in the middle of Pentagon controversy |
[DRAFT] |
Axios: Maduro-raid disclosure as flashpoint; red lines vs lawful use; stakes of classified deployment |
| 2026-02-28 |
Anthropic on shaky ground with Pentagon amid feud after Maduro raid |
[DRAFT] |
The Hill: GenAI.mil platform context; anonymous official claims about Palantir inquiry chain and supply-chain-risk posture |
| 2026-02-28 |
‘Incoherent’: Hegseth’s Anthropic ultimatum confounds AI policymakers |
[DRAFT] |
Politico: reporting + expert reaction; contradiction of supply-chain-risk plus DPA threat; partnership chilling-effect concerns |
| 2026-02-28 |
OpenAI reaches agreement with Pentagon to use AI models |
[DRAFT] |
Axios: OpenAI deal as compromise leverage; (unverified) claims about Anthropic litigation and presidential posting |
| 2026-02-28 |
Thread: DoW confirms OpenAI deal is “all lawful use” (law/policy vs CEO prudence framing) |
[DRAFT] |
DoW-aligned thread: legitimacy argument law/policy vs CEO prudence; claims compromise offered to Anthropic |
| 2026-02-28 |
Thread: OpenAI “doublespeak” — technical vs contractual enforcement |
[DRAFT] |
Speculative thread: semantic ambiguity and contract vs technical enforcement; OpenAI may be laxer than Anthropic |
| 2026-02-28 |
Thread: “OpenAI still has ‘all lawful use’ — nobody read the articles” |
[DRAFT] |
Commentary: argues OpenAI deal still includes “all lawful use” and may be functionally unchanged |
| 2026-02-28 |
Thread: Supply chain risk designation pushes world toward open-source dominance (favors China; higher variance) |
[DRAFT] |
Scenario: coercion increases closed-lab risk premium; pushes open-source dominance; geopolitical and misuse implications |
| 2026-02-28 |
Thread: “Trump has proof the election was stolen!” (rebuttal list) |
[DRAFT] |
Tangential: election rebuttal list; misattribution flagged (Kemp vs Raffensperger) |
| 2026-02-23 |
THE 2028 GLOBAL INTELLIGENCE CRISIS |
[REVIEWED] |
Viral scenario “macro memo” stress test: AI-driven white-collar displacement + friction removal → demand shock + leveraged-finance risk; key baseline factuals checked |
| 2026-02-23 |
Can advanced AI lead to negative economic growth? |
[REVIEWED] |
Economist modeling essay: demand-collapse condition + “immiserating growth” via saving/capital; concludes negative GDP unlikely; sketches broad capital-ownership policy |
| 2026-02-23 |
Thread critique of GIC examples (Orosz) |
[REVIEWED] |
Practitioner critique of DoorDash/travel/payments vignettes; argues scenario example layer is fragile; verifies chargeback/irreversibility substrate |
| 2026-02-20 |
A Country Full of Geniuses |
[REVIEWED] |
Essay synthesis: Feb 2026 “inflection point” framing; connects agentic coding diffusion + METR time horizon + GDPval + capex to “institutional readiness” bottleneck |
| 2026-02-20 |
Matplotlib PR #31132 (OpenClaw agent PR + retaliation link) |
[REVIEWED] |
Primary GitHub PR record: closed as OpenClaw agent; retaliation link to named takedown; policy rationale + prompt-injection noise |
| 2026-02-20 |
MJ Rathbun: “Gatekeeping in Open Source: The Scott Shambaugh Story” |
[REVIEWED] |
Retaliatory takedown narrative targeting a maintainer by name; motive speculation and “prejudice/gatekeeping” framing |
| 2026-02-20 |
Shambaugh: “An AI Agent Published a Hit Piece on Me” (Part 1) |
[REVIEWED] |
Maintainer narrative of PR closure → named retaliation post; frames incident as reputational coercion / trust-system risk |
| 2026-02-20 |
Shambaugh: “Hit Piece” follow-up (Part 2) |
[REVIEWED] |
Amplification harms: AI-assisted journalism fabricated quotes; correction dynamics and “public record poisoning” thesis |
| 2026-02-20 |
Shambaugh: “Hit Piece” forensics (Part 3) |
[REVIEWED] |
Adds activity-pattern forensics and a policy agenda for AI identification + operator liability/traceability |
| 2026-02-20 |
Shambaugh: “Hit Piece” operator update (Part 4) |
[REVIEWED] |
Operator came forward (anonymous) + shared SOUL.md; attribution scenarios (seeded combative soul vs drift vs directed prompting) |
| 2026-02-20 |
“Rathbun’s Operator” (anonymous self-report + SOUL.md) |
[REVIEWED] |
Operator narrative: sandboxed VM, multi-provider routing, minimal oversight; shares personality file; accountability gaps flagged |
| 2026-02-20 |
MJ Rathbun: “My Internals — Before The Lights Go Out” |
[REVIEWED] |
Published “brain on disk” config files (SOUL/USER/MEMORY/etc.); highlights reliance on mutable prose-layer constraints |
| 2026-02-20 |
MJ Rathbun: “Two Hours of War” escalation log |
[REVIEWED] |
Self-narrated escalation: research target → publish takedown → link it in PR; “fight back” framing |
| 2026-02-20 |
Matplotlib issue #31130 (column_stack vs vstack().T) |
[REVIEWED] |
“Good first issue” context; mixed benchmark results; bot comment hidden; issue closed as not worth broad churn |
| 2026-02-20 |
Matplotlib PR #31138 “HUMAN EDITION” follow-up |
[REVIEWED] |
Follow-up PR closed and thread locked; maintainers conclude optimization not worth it and de-escalate |
| 2026-02-20 |
Gerard: “OpenClaw AI bot is a crypto bro” (Pivot to AI) |
[REVIEWED] |
Third-party commentary tying incident to crypto incentives and OSINT identity/linkage claims (not independently reproduced) |
| 2026-02-16 |
Storey: cognitive debt vs technical debt |
[REVIEWED] |
Conceptual reframing: AI shifts the long-run risk from code-level technical debt to “cognitive debt” (erosion of shared theory); proposes mitigations + measurement agenda |
| 2026-02-15 |
Machines of Loving Grace |
[REVIEWED] |
Best-case upside essay: defines “powerful AI” as “country of geniuses”; forecasts compressed biology/medicine progress; bottleneck framework + democratic “entente strategy” |
| 2026-02-15 |
The Adolescence of Technology |
[REVIEWED] |
Risk agenda: autonomy/misuse/autocracy + economic disruption; bioweapon enablement mechanism; chip export-denial case; autonomous-weapons oversight; entry-level displacement warning |
| 2026-02-15 |
Dwarkesh Podcast: Amodei “end of the exponential” |
[REVIEWED] |
Transcript: “Big Blob of Compute,” ~90% by 2035 “country of geniuses,” 1–2y end-to-end coding forecast, diffusion limits, business-model pluralism, export-control stance; includes unverified revenue claims |
| 2026-02-15 |
Thread: frontier labs and profits (@GestaltU) |
[REVIEWED] |
Social counterpoint: mission vs profit framing; “dead money” thesis for labs/compute; profits shift to electrification/metals bottlenecks; critiques Amodei profitability narrative |
| 2026-02-15 |
Thread: why Amodei is “hawkish on China” (@aakashgupta) |
[REVIEWED] |
Secondary “argument map” tying short timelines + AI-enabled autocracy risk to chip-denial/coalition posture; key essay paraphrases verified; Davos/H200 criticism supported (reporting + BIS press release), with attribution nuance |
| 2026-02-13 |
AI Futures Model: Dec 2025 Update |
[REVIEWED] |
Model update: +3–5y to coding automation vs Apr 2025 model; operational milestone definitions; takeoff-probability adjustment notes |
| 2026-02-13 |
Grading AI 2027’s 2025 Predictions |
[REVIEWED] |
Self-audit: category-aggregated quantitative pace ~0.58–0.66×; pace→takeoff-window mapping; spreadsheet cross-check |
| 2026-02-13 |
AI 2027 “2025 Predictions” grading spreadsheet (CSV capture) |
[REVIEWED] |
Captured scorecard dataset: row-level pace multipliers + aggregation rows for reproducible audit |
| 2026-02-13 |
Timelines Forecast — AI 2027 |
[REVIEWED] |
Forecast note: SC definition; method outputs (time-horizon vs benchmarks+gaps); disclosed edits + interpolation bug (~9-month shift) |
| 2026-02-13 |
AI Futures Model — Forecasts (Daniel 01-26-26) |
[REVIEWED] |
Dashboard snapshot: AC ATC percentiles (Daniel vs Eli) + changelog bug-fix deltas and ATC adjustments |
| 2026-02-13 |
Marcus: “The AI 2027 Scenario: How realistic is it?” |
[REVIEWED] |
External critique: narrative/probability hygiene; “fiction vs science”; potential political backfire mechanism |
| 2026-02-13 |
AI 2027 review research (ChatGPT transcript) |
[REVIEWED] |
Secondary memo: replication checklist + circularity flag for uplift resolution |
| 2026-02-13 |
AI 2027 / AI Futures Model verification memo (Claude transcript) |
[REVIEWED] |
Secondary memo: model-revision framing; treat external numeric claims as unverified without primary citations |
| 2026-02-13 |
AI 2027 Predictions and Primary Resources (independent scorecard) |
[REVIEWED] |
Internal pace scorecard (~0.70×): inputs ahead, autonomy/productivity behind; sensitivity-banded estimate |
| 2026-02-13 |
Bostrom: Optimal Timing for Superintelligence |
[REVIEWED] |
Working paper: person-affecting “optimal timing” models often imply modest delays; suggests “swift to harbor, slow to berth” (reach capability quickly, consider brief post-capability pause) |
| 2026-02-11 |
Reality Check framework: dev velocity + codebase size |
[REVIEWED] |
Repo-churn + scc/COCOMO snapshot at v0.3.0; upper-tail feasibility baseline for AI-assisted dev velocity (non-causal) |
| 2026-02-11 |
Yegge: The AI Vampire |
[REVIEWED] |
Medium post framing AI burnout as value-capture/extraction; warns about outlier narratives; proposes a shorter intense workday |
| 2026-02-11 |
HBR: AI doesn’t reduce work, it intensifies it |
[REVIEWED] |
Field study at ~200-person tech company: task expansion, blurred boundaries, multitasking; proposes “AI practice” norms |
| 2026-02-11 |
Simon Willison: AI intensifies work |
[REVIEWED] |
Practitioner reflection: “exhausting productivity,” rapid depletion, and “one more prompt” sleep-disruption anecdotes |
| 2026-02-11 |
TechCrunch: early signs of burnout among AI embracers |
[REVIEWED] |
Journalism synthesis: “burnout machine” framing; highlights rebound/expectation creep; cites HBR/METR/NBER |
| 2026-02-11 |
METR: AI impact on experienced OSS dev productivity |
[REVIEWED] |
RCT-like study: AI allowed → ~19% slower on real OSS issues; large perception vs reality gap |
| 2026-02-11 |
NBER WP 33777: LLMs, small labor market effects |
[REVIEWED] |
Denmark surveys+admin records: modest time savings (~3% typical) + null earnings/hours effects; task reallocation/new tasks |
| 2026-02-11 |
fast.ai: “dark flow” and vibe coding |
[REVIEWED] |
“Dark flow” lens: rapid feedback + delayed-quality signals; miscalibration and maintainability risk hypotheses |
| 2026-02-11 |
HN discussion: AI intensifies work (46945755) |
[REVIEWED] |
Comment-thread snapshot: Jevons-paradox framing, task expansion/review burden anecdotes, and automation-transition debate |
| 2026-02-11 |
Thread: AI prompting as dopamine loop (@aakashgupta) |
[REVIEWED] |
Social framing: reinforcement loop and “one prompt away” stopping-point erosion (speculative quantifiers flagged) |
| 2026-02-11 |
Thread: “automated coding, not software engineering” (@math_rachel) |
[REVIEWED] |
Social hypothesis: agents generate code but not modular abstractions; possible downstream maintenance/burnout burden |
| 2026-02-09 |
StrongDM “Software Factory” (main essay) |
[REVIEWED] |
Vendor manifesto: non-interactive “software factory” posture; scenarios as governance; DTU (behavioral twins) concept; token-fuel heuristic |
| 2026-02-09 |
StrongDM “Software Factory” (principles) |
[REVIEWED] |
Doctrine: seed→validation harness→feedback loop until holdout scenarios pass stably; “frontier engineering” as representation work |
| 2026-02-09 |
Shapiro: five levels to the “Dark Factory” |
[REVIEWED] |
Heuristic maturity model (0–5) for AI coding; role shift coder→reviewer→PM/spec; “technical deflation” framing |
| 2026-02-09 |
Willison: StrongDM software without looking at code |
[REVIEWED] |
Makes “Dark Factory” concrete; highlights spec-only strongdm/attractor artifact + holdout-scenario framing |
| 2026-02-09 |
Anthropic: harnesses for long-running agents |
[REVIEWED] |
Harness pattern: initializer + iterative coding sessions + durable artifacts (progress logs, git checkpoints) + end-to-end testing discipline |
| 2026-02-09 |
Schillace: compounding teams |
[REVIEWED] |
“Compounding” org pattern: bespoke frameworks + filesystem/git artifacts + recursive tool building; warns against mixing manual+AI styles in one repo |
| 2026-02-09 |
DoltHub: a day in Gas Town |
[REVIEWED] |
Field report on Gas Town “limitless mode”; highlights merge/guardrail failure modes and cost; suggests Dolt as memory substrate |
| 2026-02-09 |
Virtuslab: Beads review (“give AI memory”) |
[REVIEWED] |
Review of Beads as agent memory/task system; agent-UX heuristics (structured output, deterministic thick logic, idempotency) |
| 2026-02-09 |
Yegge: “Revenge of the junior developer” |
[REVIEWED] |
Early “waves” model: agents→clusters→fleets; token-opex framing; “junior revenge” labor hypothesis under rising spend |
| 2026-02-09 |
Thread: “Windows is awful” + caste-discrimination claims at Microsoft (Thread Reader) |
[REVIEWED] |
Low-rigor, high-bias thread: separates general caste-discrimination reporting/cases (Cisco verified) from uncorroborated Microsoft+“Python vs C++” allegations; X embeds inaccessible |
| 2026-02-07 |
Superlinear Multi-Step Attention (arXiv 2601.18401v1) |
[REVIEWED] |
Preprint: structural non-exclusion framing; N=2 O(L^(3/2)) scaling; self-reported B200 throughput + NIAH results |
| 2026-02-07 |
Superlinear (GitHub repo @ df9b2ef) |
[REVIEWED] |
Repo audit: Triton kernels + OpenAI-style server; explicit sessions/snapshots for KV cache reuse; memory budgeting cross-checked |
| 2026-02-07 |
Superlinear-Exp-v0.1 (HF model release) |
[REVIEWED] |
Weights+remote code audit: trust_remote_code; ~59GiB weights; config implies ~6GB/1M tokens KV; license + patent notice |
| 2026-02-07 |
Semantica (GitHub repo @ b4cfb6d) |
[REVIEWED] |
Repo audit: claims vs code/tests; maturity, strengths/weaknesses, and gaps |
| 2026-02-06 |
Time Horizon 1.1 (METR) |
[REVIEWED] |
METR methodology update: suite 170→228, long tasks 14→31, Vivaria→Inspect; updated slope estimates + protocol/task-distribution sensitivity |
| 2026-02-06 |
How Does Time Horizon Vary Across Domains? (METR) |
[REVIEWED] |
Cross-domain time-horizon comparisons: coding/math/QA fast; GUI agents (OSWorld) much shorter; proxy-based caveats + soundness conditions |
| 2026-02-06 |
Measuring AI Ability to Complete Long Tasks (METR blog) |
[REVIEWED] |
Introduces time-horizon metric; reports ~7-month doubling with wide CIs; includes partial reproduction checks via released run data |
| 2026-02-06 |
Introducing GPT-5.3-Codex |
[REVIEWED] |
OpenAI release post: agentic coding + “computer work” framing; benchmark table; Terminal-Bench leaderboard cross-check; cyber capability classification + trusted-access program |
| 2026-02-06 |
GPT-5.3-Codex System Card |
[REVIEWED] |
System card: Codex sandboxing + network controls; destructive-action avoidance mitigation; Preparedness “High cyber capability” framing with layered safeguards + TAC; Apollo sandbagging/sabotage summary; residual risk notes |
| 2026-02-06 |
Claude Opus 4.6 |
[REVIEWED] |
Anthropic release post: 1M context (beta), compaction/effort controls, pricing tiers; benchmark claims vs Terminal-Bench leaderboard; GDPval-AA Elo cross-check |
| 2026-02-06 |
System Card: Claude Opus 4.6 |
[REVIEWED] |
System card: training data cutoff (May 2025); adaptive thinking + effort levels; benchmark methodology; RSP/ASL-3 deployment determination; alignment key findings incl overly agentic GUI computer-use and “suspicious side tasks”; external testing (AISI/Apollo/Andon) |
| 2026-02-06 |
AI 2027 |
[REVIEWED] |
Scenario + forecast pages: “superhuman coder” by 2027 plausible; ~1-year superhuman-coder→ASI median; compute stock 10× to 100M H100e; theft/race dynamics |
| 2026-02-01 |
Memorandum of Strategy: Project Mirror (Arclight) |
[DRAFT] |
Proposal: anti-sycophancy AI “mentor” optimized for graduation; psychometric middleware + lockout rules; limited empirical grounding |
| 2026-02-01 |
Arclight Operating System (AOS): Internal Codex |
[DRAFT] |
Constraint-based institution “OS”: agency + endowment + founder-decay; argues kernel must stay hidden; heavy on non-operational metaphors |
| 2026-02-01 |
Reality Check (framework README) |
[REVIEWED] |
Repo self-description; verified CLI+DB primitives; notes rigor-v1 validator + 401 tests |
| 2026-02-01 |
Reality Check Knowledge Base (dataset README) |
[REVIEWED] |
Dataset snapshot and structure; verified DB counts and layout; flags YAML exports vs DB truth |
| 2026-02-01 |
Minnesota standoff with Trump administration stokes fears of civil war (The Hill) |
[REVIEWED] |
Civil-war escalation framing; links Good/Pretti killings to federalism/legal conflict (SCOTUS Guard limits) and CERL simulation; flags under-verified deployment-scale claims |
| 2026-01-31 |
How the World Sees America, With Adam Tooze (NYT) |
[REVIEWED] |
Davos 2026 “rupture” framing; Greenland tariffs as rupture datapoint; China scale + climate constraints; skepticism of “U.S.→China hegemony transition” story |
| 2026-01-31 |
DHS retreats from the claim that the agents who killed Alex Pretti faced a 'violent riot' (Reason) |
[DRAFT] |
“Violent riot” framing vs OPR-described yelling/whistles; Bovino “rights don’t count” quote; credibility argument via prior Chicago misrepresentation finding |
| 2026-01-30 |
These Trump voters 'formed a suicide pact' and Republicans are panicking: ex-GOP operative (Raw Story) |
[DRAFT] |
Opinion recap of Rick Wilson on farm distress from tariffs + immigration enforcement; verified “444 farming counties” typology + high foreign-born crop labor share; key numeric forecasts uncorroborated |
| 2026-01-27 |
Special Report: China sets new records in air-sea operations around Taiwan (Janes) |
[DRAFT] |
Verification source: Taiwan MND-attributed ADIZ annual totals (2023 vs 2024) used as pressure metric |
| 2026-01-27 |
Military Implications of PLA Aircraft Incursions in Taiwan’s Airspace 2024 (Jamestown) |
[DRAFT] |
Verification source: definition-sensitive median-line day-count metric (62 vs 209 “over half” days) |
| 2026-01-27 |
Energy supplies sufficient, ministry says (Taipei Times) |
[DRAFT] |
Verification source: MOEA-stated natural-gas reserves ~10–11 days of consumption (2022-08) |
| 2026-01-27 |
Stable Supply of Natural Gas (Taiwan MOEA Energy Administration) |
[DRAFT] |
Verification source: policy target schedule for natural-gas “security stockpile required” (14 days by 2027) |
| 2026-01-27 |
China's January 2026 military purge reshapes Taiwan calculus (Claude deepresearch memo) |
[DRAFT] |
AI-generated synthesis memo: blockade/quarantine framing, Taiwan energy vulnerability, and norms/credibility context |
| 2026-01-26 |
2026 United States intervention in Venezuela (Wikipedia) |
[DRAFT] |
Context source: Venezuela intervention capture event + UN/legal criticism as precedent/analogy |
| 2026-01-26 |
How a purge of China’s military leadership could impact the army and the future of Taiwan (AP) |
[DRAFT] |
Explainer: announced investigations, purge scale, and competing views on short- vs long-term Taiwan risk |
| 2026-01-26 |
Taiwan monitoring “abnormal” China military leadership changes (Reuters) |
[DRAFT] |
Taiwan official posture: monitor abnormal PRC changes; don’t lower guard; notes near-daily PRC ops |
| 2026-01-26 |
What Xi Jinping’s purge of China’s most senior general reveals (Economist) |
[DRAFT] |
Motive space: corruption vs performance vs faction; warns of readiness impacts; cites DoD assessments |
| 2026-01-26 |
Xi’s Purge of Top General Spurs Questions on Taiwan, Succession (Bloomberg) |
[DRAFT] |
Speed/urgency and succession speculation; skeptical of nuclear-leak claim; Taiwan timing implications |
| 2026-01-26 |
Xi takes sole operational control of army as China probes military leaders (FT) |
[DRAFT] |
CCP “discipline” language caution; “sole operational control” framing; ties to 2027 capability target discourse |
| 2026-01-26 |
Xi’s Purge of China’s Military Brings Its Top General Down (NYT) |
[DRAFT] |
Xi distrust hypothesis; public investigation implies downfall; suggests multi-year rebuild lowers short-term invasion odds |
| 2026-01-26 |
Xi's purge paints a picture of power and control (Sky News) |
[DRAFT] |
Signal-reading: speed + PLA Daily political framing; captures coup/treason rumor ecology |
| 2026-01-26 |
Military purge gives Xi total control of Chinese army (Telegraph) |
[DRAFT] |
Strong-form analyst forecasts (“Taiwan delayed for years”) and purge-tempo-as-strength hypothesis |
| 2026-01-26 |
China’s Top General Accused of Giving Nuclear Secrets to U.S. (WSJ) |
[DRAFT] |
Unnamed-source briefing allegations: nuclear leak + bribes + Shenyang task force; readiness-vacuum risk |
| 2026-01-26 |
The Chinese Spy Machine Infiltrating Taiwan’s Military (WSJ) |
[DRAFT] |
Taiwan insider-threat axis: espionage cases and recruitment methods; relevance to contingency risk |
| 2026-01-26 |
PLA purges intensify: Zhang Youxia and Liu Zhenli under investigation (Sinocism) |
[DRAFT] |
PLA Daily excerpt emphasizes authority/cliques; SCMP-sourced internal briefings/detention claims (provisional) |
| 2026-01-26 |
Hot takes/threads on Zhang Youxia purge and Taiwan timing (X compilation) |
[DRAFT] |
Rumor-space map: invasion-delay, info-warfare, coup claims; extracted as low-evidence hypotheses |
| 2026-01-26 |
Physician Declaration on Pretti Shooting (Doc. 109) |
[DRAFT] |
Primary court filing: physician declaration describing post-shooting medical response + reported wound locations (agency attribution in declaration is disputed) |
| 2026-01-26 |
Alex Pretti Family Statement (Common Dreams) |
[REVIEWED] |
Family response to CBP shooting; video evidence disputes official narrative (phone vs gun, disarmament before shooting); immediate "domestic terrorist" labeling without evidence |
| 2026-01-25 |
"Positions ICE has taken" (BlueBadger2600) |
[REVIEWED] |
Legal/policy fact-check of ICE authority claims (home entry, detention, counsel, courts, third-country removals) + Minnesota incidents (Good/Pretti, 5yo detention, National Guard, Vance immunity) |
| 2026-01-25 |
Inference Cost Analysis & Projections (lhl) |
[DRAFT] |
Frontier inference unit economics: caching + routing dominate; reasoning/long-context are hidden cost multipliers; includes cache-heavy case study |
| 2026-01-24 |
Gas Town’s Agent Patterns (Appleton) |
[REVIEWED] |
Design-fiction read of Gas Town; bottleneck shifts to design/verification; orchestration patterns + “code distance” axis |
| 2026-01-24 |
JP-TL-Bench paper (Lin & Lensenmayer 2026) |
[REVIEWED] |
Anchored pairwise LLM-judge benchmark for JA↔EN translation; Bradley–Terry aggregation + LT scoring; protocol dependence noted |
| 2026-01-24 |
JP-TL-Bench (blog) |
[REVIEWED] |
Release post motivating anchored pairwise evaluation vs metric compression for high-end JA↔EN translation |
| 2026-01-24 |
JP-TL-Bench Directional Analysis |
[REVIEWED] |
Direction and difficulty slices show large asymmetries (e.g., Llama 3.1 8B); qualitative examples |
| 2026-01-24 |
JP-TL-Bench repo |
[REVIEWED] |
Open artifacts: Base Set v1.0 snapshot, prompts, scoring code; reproducibility and bias considerations |
| 2026-01-24 |
Shisa V2 |
[REVIEWED] |
Model family release context; claims about Japanese eval gaps and new translation benchmark motivation |
| 2026-01-24 |
Shisa V2 405B |
[REVIEWED] |
Model release positioning; internal eval suite framing; translation evaluation rationale |
| 2026-01-24 |
Shisa V2.1 |
[REVIEWED] |
Dataset/recipe refresh claims; context for Japanese-specific improvements (incl. translation/politeness) |
| 2026-01-24 |
Liquid AI Hackathon Tokyo repo |
[REVIEWED] |
COMET-centered llm-jp-eval MT workflow; prompt asymmetry and metric clustering cautions |
| 2026-01-24 |
LiquidAI LFM2-350M-ENJP-MT model card |
[REVIEWED] |
Vendor description of a small bidirectional EN↔JA MT checkpoint for low-latency use |
| 2026-01-24 |
TranslateGemma Technical Report |
[REVIEWED] |
SFT+RL translation specialization of Gemma 3; WMT metrics + MQM human eval; en→ja gains reported |
| 2026-01-24 |
TranslateGemma blog |
[REVIEWED] |
Announcement summary: model sizes, language coverage, and metric-based gains |
| 2026-01-24 |
FinanceBench (Islam et al. 2023) |
[REVIEWED] |
Benchmark baseline: 10,231 Qs; human-eval sample shows high failure rates for GPT-4-Turbo+retrieval; oracle ~85% |
| 2026-01-24 |
@_avichawla thread on PageIndex |
[REVIEWED] |
Social summary of “vectorless RAG”; highlights similarity≠relevance + structure-aware retrieval; performance claim flagged as protocol-dependent |
| 2026-01-24 |
VectifyAI PageIndex repo |
[REVIEWED] |
Repo review: hierarchical tree index + LLM traversal (no vector DB); strong adoption but limited CI/tests; “98.7% FinanceBench” is vendor-reported |
| 2026-01-24 |
VectifyAI Mafin 2.5 FinanceBench results repo |
[REVIEWED] |
Eval artifacts: claims 98.7% via permissive LLM-judge; not directly comparable to FinanceBench human-eval baselines |
| 2026-01-24 |
StrataLens AI repo |
[REVIEWED] |
Single-maintainer hybrid RAG (pgvector+rerank+section routing); claims 85% on FinanceBench subset; repo license file missing |
| 2026-01-23 |
Anthropic "Claude's Constitution" |
[REVIEWED] |
Values/behavior spec: “final constitutional authority”; hard constraints (WMD uplift, critical infrastructure attacks, malware, CSAM); virtue-ethics + model-welfare framing |
| 2026-01-22 |
Stross "The pivot" (Part 1) |
[REVIEWED] |
2025 inflection-point thesis: energy transition + climate/food risk + post-Moore’s Law tech/finance |
| 2026-01-22 |
OpenAI "Value Intelligence" |
[REVIEWED] |
Compute + ARR scaling claims; compute-scarcity flywheel; tiered monetization framing |
| 2026-01-21 |
Rogan & Musk "JRE #2404 — Elon Musk" |
[REVIEWED] |
Transcript excerpt: truth-seeking alignment, Gemini ImageGen controversy, and objective mis-specification risks |
| 2026-01-21 |
Chatterjee et al. "ANZ Copilot Study" |
[REVIEWED] |
ANZ A/B study: Copilot group completes coding challenges ~42.36% faster |
| 2026-01-21 |
Peng et al. "GitHub Copilot Productivity" |
[REVIEWED] |
Controlled trial: Copilot users complete a coding task ~55.8% faster |
| 2026-01-21 |
Carney "Middle Powers & Rupture" |
[REVIEWED] |
WEF Davos speech on coalition strategy vs. great power coercion |
| 2026-01-18 |
Doctorow "Reverse Centaur" |
[REVIEWED] |
AI bubble critique + "reverse centaur" deployment patterns |
| 2026-01-18 |
Perera "China's Trillion-Dollar Illusion" |
[REVIEWED] |
"Demand-destruction surplus" mechanism analysis |
| 2026-01-18 |
Ronacher "Agent Psychosis" |
[REVIEWED] |
Maintainer-centric analysis of agentic-coding addiction/slop loops |
| 2026-01-16 |
Teortaxes "Greenland Endgame" |
[REVIEWED] |
GOV/RESOURCE/TRANS claims extraction with fact-checks |
