Microsoft cut internal Claude spend by more than 33 % and reduced the per‑employee token budget from $100k/month to $10k/month, forcing Copilot to auto‑route to cheaper models.
META halved active Claude users from 60 000 to 30 000 and replaced Claude with “Muse Code”.
A user reported using approximately 4 % of the weekly Max plan tokens.
The open‑source “Kolibri” model has 78 B parameters, 3.46 B active parameters, and supports up to 1 M token context; it is built in Europe and released under Apache 2.0.
Frontier AI models are now faster and more accurate than junior accountants on medium‑length accounting tasks, a reversal from 18 months ago when they scored well below humans.
In early October 2026 I posted on X about a rapid escalation of AI capability and its organizational fallout, noting a 500 000‑fold improvement over the previous result and a 2^167‑fold improvement over the original OpenAI result.
A 500 thousand fold improvement over the previous result and a 2 ^ 167 fold improvement over the original OpenAI result.
This result was a true community effort that combined techniques from a number of contributors.
Organizations are a form of narrow superintelligence.
the failure to deeply consider how to make AI work together with existing organizational superintelligences is a big reason why we have not yet seen any large impact of AI on scientific discovery or economic productivity despite high levels of AI ability.
As a researcher who did early work on the productivity impacts of AI chatbots using RCTs, I’d note a lack of similar studies since the dawn of true agents last fall.
The few attempts to measure similar things (METR Long Tasks, GDPval) all became saturated earlier this year as AIs began to do days of work at a high quality level.
GPT-6 and Intelligent UI, now rolling out in ChatGPT for everyone.
I think most people who have been closely watching AI for the past few years are likely surprised by how relatively rare terrible AI incidents have been, given a billion users overall and hundreds of millions of corporate users.
When people talk about AI easily replacing workers, I think about this ethnography of copier repair technicians in the 1990s.