This is a curated reading list, last updated September 11, 2026, covering the state of open AI models across strategy, safety, economics, and US-China competition. It is not a summary. It is a structured syllabus, and it assumes you want to understand the field seriously. The foundation section alone includes Irene Solaiman's 2023 arxiv paper on release gradients, Sayash Kapoor and Rishi Bommasani's 2024 work on societal impact of open foundation models, and Shayne Longpre's July 2024 analysis of the collapse of the AI data commons.
The list's most useful function is mapping the open-versus-closed debate onto real economic and geopolitical stakes. Christian Catalini's August 2026 piece draws on IP history to argue open models capture value as complements to closed ones. Nathan Lambert's interconnects posts track China's escalation through specific model releases: Kimi K3 and GLM-5.2. The ATOM Report from April 2026 provides adoption data split by US and Chinese models. These are not opinion pieces. They are data-grounded positions on who is winning and why.
The list is updated over time and accepts reader submissions. The safety subsection is worth reading in full even if you skip everything else: it puts Florian Brand's June 2026 argument that closed model guardrails fail constantly against the theoretical risks of open weights, which have not materialized at scale. That tension is the core policy fight happening right now, and this list gives you the primary sources to engage with it directly.
[READ ORIGINAL →]