The period from the launch of ChatGPT on November 30, 2022 through the end of 2023 is the thinnest stretch of the wiki's sources/ record relative to its significance. This page records what the wiki holds for that period and what it does not, as of July 31, 2026.
Source density by year
sources/ pages, counted by their date: frontmatter:
| Year | Pages | Year | Pages |
|---|---|---|---|
| ≤2011 | 7 | 2020 | 2 |
| 2012 | 1 | 2021 | 6 |
| 2013 | 1 | 2022 | 10 |
| 2014 | 4 | 2023 | 24 |
| 2015 | 2 | 2024 | 48 |
| 2016 | 6 | 2025 | 161 |
| 2017 | 4 | 2026 | 286 |
| 2018–2019 | 5 |
The distribution reflects when the wiki was built rather than when the events mattered: 2025 and 2026 together account for roughly three-quarters of all source pages. Two of the 2022 and 2023 entries were added in this cycle.
What exists
Model and capability record. ChatGPT: Optimizing Language Models for Dialogue (OpenAI, November 30, 2022) (the launch post, added July 31, 2026), Training Language Models to Follow Instructions with Human Feedback (InstructGPT) (the RLHF method ChatGPT was built on), Training Compute-Optimal Large Language Models (Chinchilla) (the compute-optimal scaling result), GPT-4 Technical Report (OpenAI, arXiv 2303.08774, March 2023) and GPT-4 System Card (OpenAI, March 2023) (added July 31, 2026 and earlier), Sparks of Artificial General Intelligence: Early experiments with GPT-4 (on an early unreleased GPT-4 build), Textbooks Are All You Need (Gunasekar et al., Microsoft Research, 2023), Weak-to-Strong Generalization (Burns et al., OpenAI, 2023), and the benchmarks GPQA: A Graduate-Level Google-Proof Q&A Benchmark and SWE-bench: Can Language Models Resolve Real-World GitHub Issues?. The model pages GPT-4 Family (OpenAI) and GPT-5 Family (OpenAI) carry the product lineage.
Safety and alignment. Constitutional AI: Harmlessness from AI Feedback (2022), Core Views on AI Safety: When, Why, What, and How, AI Control: Improving Safety Despite Intentional Subversion (Greenblatt, Shlegeris, Sachan & Roger, 2023), FLI — Pause Giant AI Experiments: An Open Letter (the March 2023 pause letter) and Statement on AI Risk (CAIS) (the May 2023 one-sentence statement) — the two collective statements the task list flagged are both present.
Governance instruments. OSTP Blueprint for an AI Bill of Rights (October 2022) (2022), Executive Order 14110 — Safe, Secure, and Trustworthy AI (October 2023), NIST AI Risk Management Framework (AI RMF 1.0) (January 2023), The Bletchley Declaration (AI Safety Summit, 1–2 November 2023) and G7 Hiroshima Code of Conduct for Advanced AI (2023) (both late 2023), China — Provisions on the Administration of Deep Synthesis Internet Information Services (2022) and China — Interim Measures for the Management of Generative AI Services (2023), Canada AIDA (Bill C-27, Part 3) — Source Summary, Regulation (EU) 2022/1925 — Digital Markets Act (primary text), and DoD Directive 3000.09 — Autonomy in Weapon Systems (source summary).
Litigation and economics. NYT v. OpenAI and Microsoft — Complaint (Dec 2023) with its tracking page NYT v. Microsoft, OpenAI et al.; Generative AI at Work; and the period commentary in Attenuating Innovation (AI) — Ben Thompson (Stratechery, October 2023), The Coming Wave — Suleyman and Bhaskar (2023), Digital Empires — Anu Bradford (2023), Misinformation on Misinformation: Conceptual and Methodological Challenges — Altay, Berriche, Acerbi (Social Media + Society, 2023) and How a Fervent Belief Split Silicon Valley—and Fueled the Blowup at OpenAI — Robert McMillan, Deepa Seetharaman (WSJ, November 22 2023).
What is missing
Landmarks of the period with no dedicated page under any slug:
| Gap | What would fill it | Why it matters |
|---|---|---|
| Claude 1 / Claude 2 (March and July 2023) | An Anthropic model page or launch-announcement source | The wiki's models/ record of Anthropic begins with the 4.x family; the origin of the product line is absent, leaving the Constitutional AI paper without a deployed artifact to attach to. |
| LLaMA 1 (February 2023) and the open-weights wave | The LLaMA paper as a sources/ page plus a models/llama-1 page | models/ holds only Llama 3 (Meta) and Llama 4 (Meta). LLaMA 1's leak and the derivative wave it produced are the origin of the open-weights policy debate the wiki carries extensively for 2025–2026. |
| Google Bard (February–March 2023) and the code-red response | A model or product page, and the launch announcement | The competitive response that made the period a race rather than a single launch is unrecorded; Google DeepMind begins its narrative later. |
| The Italian Garante's ChatGPT suspension (March–April 2023) | A legislation or enforcement page; Garante per la protezione dei dati personali (Italy) exists and mentions the regulator but not this action | The first regulatory suspension of a frontier product in a major market, and the template for GDPR-based AI enforcement. |
| The first generative-AI copyright suits (early 2023) | litigation/ pages for Andersen v. Stability AI, Getty Images v. Stability AI, and Tremblay v. OpenAI | litigation/ runs to more than 30 cases but begins effectively with the December 2023 NYT complaint; the wave that preceded it by ten months is absent, so the doctrinal starting point of the fair-use argument is missing. |
| The Bletchley run-up (mid-2023) | The UK's convening documents and the July 2023 White House voluntary commitments | The Bletchley Declaration (AI Safety Summit, 1–2 November 2023) records the outcome without the process that produced it. |
| GPT-3.5 / the pre-ChatGPT deployment record | No single artifact; would be covered by the LLaMA and Claude gaps above | Noted for completeness: the ChatGPT launch post states only that GPT-3.5 "finished training in early 2022." |
Assessment
Of the nine landmarks the request named, four were already covered — the FLI pause letter, the CAIS statement, the Bletchley Declaration, and (as of this cycle) ChatGPT and GPT-4 — and five are genuinely absent: Claude 1, LLaMA 1, Bard, the Italy suspension, and the first copyright suits. The measured thinness is real but narrower than the raw counts suggest: the governance record for 2022–2023 is reasonably complete, while the model and litigation records for the period are not. The five gaps were filed as INGEST- tasks in the queue on July 31, 2026 rather than filled here, since each requires fetching and authenticity-verifying a primary source.
Relationships
- related: ChatGPT: Optimizing Language Models for Dialogue (OpenAI, November 30, 2022), GPT-4 Technical Report (OpenAI, arXiv 2303.08774, March 2023) — the two anchors added to the period this cycle.
- related: ImageNet Classification with Deep Convolutional Neural Networks (Krizhevsky, Sutskever and Hinton, NeurIPS 2012), Mastering the Game of Go with Deep Neural Networks and Tree Search (Silver et al., Nature, January 2016) — the earlier deep-learning-era anchors added in the same pass.
- related: AI Copyright — the doctrinal thread whose 2023 starting point is the largest single gap.
- related: Open-Weight Frontier Models — the debate whose origin event, the LLaMA 1 release, has no page.