# The Discontinuity Thesis > A Sequence of Seven Essays on Why Postwar Capitalism Ends Author: Ben Luong Version: v1.1.2 final Publication date: May 2026 References: 64 Canonical HTML: https://thesis.copecheck.com/ Plain text: https://thesis.copecheck.com/full.txt Structured JSON: https://thesis.copecheck.com/data.json Open Knowledge Format bundle: https://thesis.copecheck.com/okf/ Open Knowledge Format browser: https://thesis.copecheck.com/okf-browser/ Open Knowledge Format zip: https://thesis.copecheck.com/okf.zip PDF: https://thesis.copecheck.com/assets/The_Discontinuity_Thesis_v1_1_2.pdf ## Purpose This is the CopeCheck one-page web edition of The Discontinuity Thesis. The complete thesis text is present in the HTML and is also exposed as plain text and JSON for agents. ## Appendix III live register Status: live_evidence_active Updated: 2026-07-16T00:00:00Z Summary: Appendix III is wired as a live evidence register. The baseline document remains fixed at v1.1.2; scanner additions can be appended here without rebuilding the thesis page. - Kimi K3 — HuggingFace Model Card (Moonshot AI) URL: https://huggingface.co/moonshotai/Kimi-K3 Publisher: Moonshot AI Category: capabilities Claim: Kimi K3’s official HuggingFace model card documents a 2.8T-parameter MoE model with 104B active parameters, scaling to 16-of-896 experts per token via a novel Stable LatentMoE framework — yielding ~2.5× efficiency improvement over Kimi K2. Benchmark scores against frontier closed models: GPQA Diamond 93.5 (vs GPT-5.6 Sol 94.1, Claude Fable 5 92.6); DeepSWE 67.5 (vs GPT-5.6 Sol 73.0, Claude Opus 4.8 59.0); BrowseComp 91.2 (vs GPT-5.6 Sol 90.4, Claude Fable 5 88.0); MCPMark-Verified 94.5 (vs GPT-5.6 Sol 92.9, Claude Fable 5 87.4); SWE-Marathon 42.0 (SOTA; GPT-5.6 Sol 39.0, Claude Fable 5 35.0); Terminal-Bench 2.1 88.3 (vs GPT-5.6 Sol 88.8); OSWorld-Verified 84.8 (vs Claude Fable 5 85.0); JobBench 54.3 (vs Claude Fable 5 57.4, GPT-5.6 Sol 45.4); AutomationBench 30.8 (SOTA across all listed models). Architecture: 93 layers (1 dense + 92 MoE), 69 KDA + 24 Gated MLA attention layers, 7168-dim attention, 96 heads, vocabulary 160K tokens, native MXFP4 weight quantization with MXFP8 activations trained from SFT stage. Open weights released under Kimi K3 License. Relevance: Appendix III — capabilities: official technical specification and benchmark data for Kimi K3; cross-validates frontier-level autonomous performance (SWE-Marathon SOTA, AutomationBench SOTA, MCPMark-Verified SOTA) as open-weight model; JobBench score directly relevant to job displacement thesis - FRED Real-Time Population Survey — GenAI Work Penetration URL: https://fred.stlouisfed.org/categories/8 Publisher: Federal Reserve Bank of St. Louis / Harvard Category: confirming_mechanism Claim: The Federal Reserve Real-Time Population Survey records that GenAI now assists 1.7% of total US work hours (up from 1.4% in Q4 2024) — the most direct available measure of displacement rather than adoption intent. Separately, 37.4% of employed adults report using GenAI for work as of Q3 2025 (up from 33.3% a year prior), and 47% of US employees say their organisation has formally integrated AI tools for productivity or efficiency as of Q2 2026 (up from 41% the prior quarter). As a self-reported survey, 1.7% is almost certainly a systematic undercount: workers routinely underestimate AI assistance embedded in their own outputs, and the measure captures only conscious use. The Federal Reserve publishing 137 quarterly series tracking AI work penetration is itself a confirmation signal — central banks do not build longitudinal surveillance infrastructure for phenomena they expect to be transient. Relevance: Appendix III — confirming mechanism: Federal Reserve longitudinal data quantifies GenAI work-hour penetration at 1.7% of total US hours (rising), with occupation-level disaggregation identifying legal, office admin, and computer/math roles as fastest-displacement cohorts; central bank surveillance infrastructure confirms the displacement effect is institutionally acknowledged as real and durable - Kimi K3 — Moonshot AI URL: https://www.kimi.com/blog/kimi-k3 Publisher: Moonshot AI Category: confirming_mechanism Claim: Kimi K3 delivers two headline displacement compressions: (1) Reproduced I-Love-Q astrophysics universal relations autonomously in ~2 hours — cross-validating 20+ papers, 300+ equations of state, 3,000+ lines of Python, and generating an interactive HTML dashboard — work the team describes as 'what would typically require one to two weeks of work by an experienced researcher.' (2) Edited its own teaser video autonomously from 56 source clips (clip selection, beat sync, audio processing, multiple revision rounds) in a domain that 'typically takes an experienced editor one to two working days, or a beginner three to five.' Supporting signals: designed a complete chip (4mm², 100MHz, 8,700 tokens/s decode throughput in simulation) in a single 48-hour autonomous EDA run using open-source tools — 'a chip built by a model, for a model'; built MiniTriton, a Triton-like GPU compiler with its own IR, optimization passes, and PTX codegen, rivalling Triton's extensively optimised stack; an early K3 version 'handled the majority of the team's kernel optimization works' during late development, recursive self-improvement in practice. Open weights release July 27 brings frontier-level displacement capability to zero marginal cost. API at $3/$15 per 1M tokens. Kimi models have held the upper bound of open-model sizes for 9 of the past 12 months. Relevance: Appendix III — confirming mechanism: research time compression (1-2 weeks → 2 hours), autonomous video editing (1-2 days → hours), 48-hour autonomous chip design, recursive self-improvement in own development, and open-weight frontier capability redistributed at zero cost from July 27 - GPT-5.6 — OpenAI URL: https://openai.com/index/gpt-5-6/ Publisher: OpenAI Category: confirming_mechanism Claim: GPT-5.6 Sol achieves new SOTA on Agents' Last Exam (52.7%), Coding Agent Index (80), and BrowseComp (92.2%). The RSI Index — measuring recursive self-improvement capability — scores 57.9% vs 41.7% for GPT-5.5; AI now writes a majority of research code inside OpenAI itself. Knowledge work outputs across presentations, financial models, and legal documents are described by early adopters as ready for production without human polish. At $5/$30 per 1M tokens, frontier-level professional displacement is commercially accessible at scale. The RSI signal is the landmark: the model can accelerate its own successor’s development, compressing the timeline further. Relevance: Appendix III — confirming mechanism: RSI Index at 57.9% (was 41.7% for GPT-5.5) and AI-majority research code authorship inside OpenAI — direct evidence of recursive capability compression accelerating the discontinuity timeline - Seedream 5.0 Pro — ByteDance URL: https://seed.bytedance.com/en/blog/beyond-generation-it-understands-design-introducing-seedream-5-0-pro Publisher: ByteDance Category: confirming_mechanism Claim: ByteDance’s Seedream 5.0 Pro crosses the threshold where creative/design work is no longer ‘hard to generate, easy to verify.’ The model generates complex infographics, multi-chart layouts, and structured design documents; performs pixel-level interactive editing from sketches; produces photorealistic portraits with lighting and motion blur; and handles 10+ languages natively. The creative sector’s traditional comparative advantage — that humans could at least verify quality even if AI could generate — no longer holds. Both generation and verification are now AI-accessible, accelerating displacement in graphic design, visual communication, and creative production roles. Relevance: Appendix III — confirming mechanism: capability threshold crossing in creative/design AI — both generation and verification now AI-accessible, eliminating human comparative advantage in graphic design and visual communication - Ireland Dept of Finance — ICT Employment by Age Cohort URL: https://www.gov.ie/en/organisation/department-of-finance/ Publisher: Ireland Department of Finance Category: confirming_mechanism Claim: Ireland’s Department of Finance data shows ICT employment among workers aged 15–29 fell 32% while workers aged 30–59 grew 6.9% and 60+ grew 10.5%. The sharp youth-skewed divergence is direct empirical evidence for the discontinuity mechanism: AI handles the entry-level execution and production tasks that junior workers provide, while experienced workers who direct, manage, and integrate AI outputs remain in demand. This is not a generational lag in skills — it’s structural displacement at the point of entry, consistent with experience creep and the closing of the junior-to-senior career ladder. Relevance: Appendix III — confirming mechanism: age-cohort displacement in ICT — youth ICT employment −32% vs experienced workers growing, Ireland DoF data - Federal Reserve Bank of New York, Liberty Street Economics (May 2026): “Do Job Postings Show Early Labor-Market Effects of AI?” URL: https://libertystreeteconomics.newyorkfed.org/2026/05/do-job-postings-show-early-labor-market-effects-of-ai/ Publisher: Federal Reserve Bank of New York Category: disconfirming Claim: Using Lightcast job postings data combined with Anthropic-derived AI exposure measures, NY Fed researchers report three findings: (1) the relative decline in vacancies for AI-exposed occupations began before the release of ChatGPT in late 2022; (2) no divergence between junior and senior postings within highly exposed occupations; (3) fewer than 10% of workers and vacancies sit in occupations with high measured AI exposure. The authors conclude these patterns make it difficult to attribute the recent slowdown in entry-level hiring to AI alone. Relevance: Appendix III — audit register: attribution disconfirmation on junior/senior vacancy split - AutoScout24 scales engineering with AI-powered workflows URL: https://openai.com/index/autoscout24/ Publisher: OpenAI Category: deployment Claim: OpenAI reports that AutoScout24 rolled out ChatGPT to roughly 2,000 employees and Codex to roughly 1,000 builder employees, with selected projects compressed from 2-3 weeks to 2-3 days. Relevance: Appendix III, section four: enterprise deployment evidence - Chatham Financial trade validation compressed from 30 minutes to under 4 URL: https://www.linkedin.com/company/openai/ Publisher: OpenAI for Business / Chatham Financial Category: deployment Claim: OpenAI for Business and Chatham Financial described a GPT-5.5-Codex workflow that reduced trade validation from roughly 30 minutes to under 4 minutes, with real-time compliance monitoring for 160+ registered employees and audit-ready workflow outputs. Relevance: Appendix III, section four: enterprise deployment evidence - Agents, robots, and us: how AI reshapes work and skills in Europe URL: https://www.mckinsey.com/mgi/our-research/agents-robots-and-us-how-ai-reshapes-work-and-skills-in-europe Publisher: McKinsey Global Institute Category: labour_market Claim: McKinsey Global Institute estimates that 58% of current work hours across ten European countries are technically automatable with existing technologies, including 44% by agents and 14% by robots. Relevance: Appendix III, sections five to seven: labour-market evidence and deployment continuation - Working with AI: measuring the occupational implications of generative AI URL: https://www.microsoft.com/en-us/research/publication/working-with-ai-measuring-the-occupational-implications-of-generative-ai/ Publisher: Microsoft Research Category: labour_market Claim: Microsoft Research analysed 200,000 anonymised Bing Copilot conversations and mapped generative AI applicability across occupations, with high exposure concentrated in communication, analysis, writing, sales, and knowledge-work roles. Relevance: Appendix III, sections five to seven: labour-market evidence and provider framing - Trinity College Dublin and Microsoft Ireland Research Shows a Widening AI Maturity Gap Between SMEs and Large Organisations URL: https://news.microsoft.com/source/emea/features/trinity-college-dublin-and-microsoft-ireland-research-shows-a-widening-ai-maturity-gap-between-smes-and-large-organisations/ Publisher: Microsoft Source EMEA / Trinity College Dublin Category: labour_market Claim: The AI Economy Ireland 2026 report says 92% of Irish organisations use or plan to use AI, but only 10% describe deployment as advanced or frontier-level; large organisations are more than twice as likely as SMEs to save 2+ hours per week per employee, while formal AI policy is associated with 10x higher rates of major productivity gains. Relevance: Appendix III, sections five to seven: labour-market evidence, organisational readiness, and deployment continuation - Accenture Ireland: Generating Impact — Turning Frontier AI Capabilities into Frontline Productivity and Growth in Ireland URL: https://www.accenture.com/content/dam/accenture/final/accenture-com/document-fy26/q3/Generating-Impact-Ireland.pdf Publisher: Accenture Ireland Category: consultancy_cope Claim: Accenture reports that 82% of Irish working hours are now ‘AI-reinventable’ (up from 42% in 2024), that AI is already being used for tasks accounting for 20% of working hours, and that 39% of Irish employees expect their job to be unrecognisable or disappear completely by end of the decade. Entry-level hiring demand expectations have deteriorated sharply: share of executives expecting increased entry-level demand fell from 49% to 33%, while those expecting reduced demand rose from 21% to 37%. Writing and editing declined across 51 Irish occupations 2023–2025. Relevance: Appendix III, section six: consultancy cope framing as evidence signal — the firms selling AI transformation are now publishing displacement data inside productivity narratives - Singular Bank helps bankers move fast with ChatGPT and Codex URL: https://openai.com/index/singular-bank/ Publisher: OpenAI Category: deployment Claim: OpenAI reports that Singular Bank built Singularity, an internal assistant powered by ChatGPT and Codex that analyzes portfolios, recommends next actions in real time, prepares meetings, and generates compliant follow-up communications. The published results are 60-90 minutes saved per banker per day and less than one minute of client-meeting preparation. Relevance: Appendix III, section four: financial-sector deployment evidence — private-banking workflows compressed by AI assistance - Databricks brings GPT-5.5 to enterprise agent workflows URL: https://openai.com/index/databricks Publisher: OpenAI Category: benchmark Claim: Databricks uses GPT-5.5 for enterprise agent workflows after the model set a new state of the art on the OfficeQA Pro benchmark. Relevance: Appendix III, section one: model and benchmark capability evidence - A new personal finance experience in ChatGPT URL: https://openai.com/index/personal-finance-chatgpt Publisher: OpenAI Category: vendor Claim: Preview a new personal finance experience in ChatGPT for Pro users in the U.S. Securely connect your financial accounts and get AI-powered insights and guidance grounded in your financial context, goals, and priorities. Relevance: Appendix III, section two: vendor threshold and platform capability evidence - Sea's View on the Future of Agentic Software Development with Codex URL: https://openai.com/index/sea-david-chen Publisher: OpenAI Category: deployment Claim: Sea Limited's CPO explains why the company is deploying Codex across engineering teams to accelerate AI-native software development in Asia. Relevance: Appendix III, section four: enterprise deployment evidence - Work with Codex from anywhere URL: https://openai.com/index/work-with-codex-from-anywhere Publisher: OpenAI Category: vendor Claim: Use Codex anywhere with the ChatGPT mobile app. Monitor, steer, and approve coding tasks in real time across devices and remote environments. Relevance: Appendix III, section two: vendor threshold and platform capability evidence - Helping ChatGPT better recognize context in sensitive conversations URL: https://openai.com/index/chatgpt-recognize-context-in-sensitive-conversations Publisher: OpenAI Category: vendor Claim: Learn how new ChatGPT safety updates improve context awareness in sensitive conversations, helping detect risk over time and respond more safely. Relevance: Appendix III, section two: vendor threshold and platform capability evidence - Building a safe, effective sandbox to enable Codex on Windows URL: https://openai.com/index/building-codex-windows-sandbox Publisher: OpenAI Category: vendor Claim: Learn how OpenAI built a secure sandbox for Codex on Windows, enabling safe, efficient coding agents with controlled file access and network restrictions. Relevance: Appendix III, section two: vendor threshold and platform capability evidence - Our response to the TanStack npm supply chain attack URL: https://openai.com/index/our-response-to-the-tanstack-npm-supply-chain-attack Publisher: OpenAI Category: vendor Claim: OpenAI details its response to the TanStack “Mini Shai-Hulud” supply chain attack, outlines protections taken to secure systems and signing certificates, and explains why macOS users must update OpenAI apps by June 12, 2026. Learn what happened, what was affected, and how OpenAI is strengthening defenses against evolving software supply chain threats. Relevance: Appendix III, section two: vendor threshold and platform capability evidence - What Parameter Golf taught us about AI-assisted research URL: https://openai.com/index/what-parameter-golf-taught-us Publisher: OpenAI Category: benchmark Claim: Parameter Golf brought together 1,000+ participants and 2,000+ submissions to explore AI-assisted machine learning research, coding agents, quantization, and novel model design under strict constraints. Relevance: Appendix III, section one: model and benchmark capability evidence - How NVIDIA engineers and researchers build with Codex URL: https://openai.com/index/nvidia Publisher: OpenAI Category: benchmark Claim: Teams use Codex with GPT-5.5 to ship production systems and turn research ideas into runnable experiments. Relevance: Appendix III, section one: model and benchmark capability evidence - OpenAI Campus Network: Student club interest form URL: https://openai.com/index/openai-campus-network-student-club-interest-form Publisher: OpenAI Category: vendor Claim: Join the OpenAI Campus Network—connect student clubs worldwide, access AI tools, host events, and build an AI-powered campus community. Relevance: Appendix III, section two: vendor threshold and platform capability evidence ## Use Use `/full.txt` when a single plain-text copy is needed. Use `/data.json` when headings, blocks, source PDF metadata, and the live Appendix III register are needed. Use `/okf/` or `/okf.zip` when an agent wants the same knowledge as linked Markdown concepts with YAML frontmatter. References extracted from the canonical PDF footnotes are available in `/data.json` under `thesis.references` and appended to `/full.txt`.