Why open weights are not open source, in a new essay on AI’s category problem
aronchick · x · 2026-07-28
- The post points to a long essay arguing that people keep calling open weights and open models “open source,” even though the author says that is misleading.
- The article revisits why the term open source was coined in 1998: to focus on source access rather than price.
- It contrasts true source availability with “free binaries” style distribution, where users can run something but cannot inspect, rebuild, or carry it forward.
- The author argues that open-weight AI is not the same thing as open source and frames the debate as a category error around what users actually receive when they download a model.
Related event: Open Weights vs. Open Source: Clarifying the AI Misnomer(3 posts)→
More from Companies & People
- Bolt signs open-weights letter as it doubles down on open source — damianplayer · 2026-07-28
- Palantir CEO Alex Karp to discuss sovereign AI live on Fox Business — eliano · 2026-07-28
- JMU and CGI launch an NSF-funded AI fellowship for rural career changers — ArtificialOther · 2026-07-28
- YC expands AI startup support with $25k compute credits and a GPU cluster — ycombinator · 2026-07-28
- All-In clip says companies adopting AI tend to grow faster — thefluffyscrum · 2026-07-28
- Nvidia mentions RicursiveAI in its DAC 2026 opening keynote — Azaliamirh · 2026-07-28