[jeremyjh]

> Opus was the reason I paid for GitHub copilot, but they had the pricing model completely wrong. I could assign copilot to a substantial issue using Opus and have it handle 30 minutes of work with many subagents with iterative testing. With 300 "premium requests" a month I could have copilot do substantial work for 3 premium requests per issue. It was very clear that this was unsustainable for Microsoft to pay for, so I expected change to come.

Yes I was hearing that a lot the past few weeks. Its pretty clear what happened:

Anthropic demand soared from OpenClaw, but they were already over-sold. Cowork shipped, Hegseth flexed and a lot of people and entire orgs moved from ChatGPT to Claude in the space of a month. They couldn't handle the demand - they quantized their models, dropped effective usage limits and made all kinds of tweaks in Claude Code to reduce token burn.

Lots of customers fled to Codex and they got crunched as well. Some people noticed Copilot was still selling dollars for nickles and mentioned it to other people.

The only question I care about: Is z.ai next? GLM 5.1 is simply where its at right now. It is not the best model, but it is much better than Sonnet at 1/5 the cost.