Newsletter Monday, October 5, 2026
Daily update on the AI tools we track. Headings come from public sources (X, GitHub, Hacker News); the Dutch commentary and summary are machine-generated. Each heading links to our own news page on that topic; the source link is next to it.
In brief
Aleph Alpha has released the open-weights model Kolibri-1. Additionally, the Allen Institute launched Olmo-core 3, focused on training trillion-parameter Mixture-of-Experts models. Finally, recent LLM updates show a pattern of incremental improvements in speed and intelligence combined with new pricing models.
Models & pricing
- LLM updates are starting to feel like iPhone releases. New model: slightly faster slightly smarter new pricing everyone argues for 3 days · source: X · 2026-09-29
The post discusses a new LLM model that is slightly faster and smarter with a new pricing model, which is relevant to indie developers because it can impact their costs and performance.
Repos, releases & updates
- AI BRIEF | Oct 4, 2026 Open weights and the data-center fight @ryan_eminhizer at X Studios MODEL RELEASES Aleph Alpha dropped Kolibri-1 on O · source: X · 2026-10-04
The article reports the release of Kolibri-1 by Aleph Alpha, an open-weights model, which could be interesting for indie developers to use without large data centers. - 🤖 AI NEWS ROUNDUP — October 3, 2026 1️⃣ ALLEN INSTITUTE RELEASES OLMO-CORE 3 FOR TRILLION-PARAMETER MoE TRAINING Ai2 shipped Olmo-core 3, an · source: X · 2026-10-03
The headline reports that the Allen Institute has released Olmo-core 3 for trillion-parameter MoE training, which is relevant to indie developers who want to train or use large models.