Yoav Goldberg: server-side session APIs will make LLMs opaque — as bad as closed weights
yoavgo · x · 2026-11-19
Yoav Goldberg highlights an overlooked dimension of "openness in AI": API openness.
- Traditional LLM APIs map transparently to the messages actually sent, giving users visibility and fine-grained control.
- Anthropic and OpenAI are shifting to APIs where session state lives server-side; he predicts the old APIs will soon be deprecated, leaving users talking to an opaque system rather than an LLM.
- Consequence: far less visibility and intervention points, making research on agentic harnesses and context management practically impossible.
He argues this is as important as open weights — maybe more — calls for loud resistance, and asks how to organize around the cause.
More from AGI Musings
- Elicit on agent swarms: pharma winners of 2036 will robustly deploy them without going rogue — jungofthewon · 2026-11-19
- Pro vs. con arguments on AI "handover" rest on different moral foundations — 1a3orn · 2026-11-19
- Google Brain researcher warns LLM literature audits lack judgment, risk witch hunts — rbhar90 · 2026-11-19
- Follow-up: a transformer has no memory, self, or cognition, dev argues — gerardsans · 2026-11-19
- The site-builder era is repeating: people will hire humans to use AI app builders, says dev — sull · 2026-11-19
- Blogger goes contrarian: AI isn't currently moving toward higher general intelligence — LucaAmb · 2026-11-19