Comment by aroman
5 days ago
This makes me think they really are quite capacity constrained at the moment.
I had assumed they were primarily limiting it to entice people to upgrade, but I feel like these limits are so low and so temporary (especially over July 4th weekend in the US) that people will barely get a chance to get "used to it" and then think: "man, I can't live without this, I'll pay for API pricing".
That's strange, because they were seemingly way less capacity constrained lately, raised limits and removed the peak hours usage. It's crazy to think that even spending $1.25 billion a month to rent GPUs from SpaceX didn't do much to improve the situation.
I don't know, I feel like for a few weeks before the SpaceX datacenter, I was just constantly checking my weekly limits. And now after that miraculously, I rarely even come close to hitting my weekly limits. and I still have 5-7 claudes open a day (defaulting to Opus 4.8 xhigh, sometimes ultracode).
So I feel that the additional datacenter caused them to just ease up a bit. But demand is also insane, so who knows...
Yea, i'm on x20 and while it has been up and down in terms of token-usage-UX, i feel like its the best its ever been. Context: I entirely use Opus 4.8 fwiw.
Now is that because 4.8 is nerfed compared to 4.6 and thus more token efficient? No idea. I just know on x20 with a pretty plain workflow i struggle to use my tokens every week.
If it's API pricing, I'm going to ditch Claude Code and switch to a harness that can jump between GLM and Claude Code.
Cheap pricing is why I use Claude Code. The minute they fumble that, I'm using Chinese models for 90% of the work.
Yeah. Their cheap subscriptions are the only reason to keep using them. If they ruin the plans there's nothing holding us back anymore.
I dont believe they can afford to switch to API pricing. Everyone will leave. I am easily spending the equivalent of 1000 dollars a day on tokens with two max subscriptions. that about 400 dollars a month. Thats acceptable for my position. But thats like 30k per month. Totally not viable.
how can a harness switch between glm (a model) and claude code (another harness)?
I've been doing this for ages - you just spin up harness B as a subprocess/tool call from harness A. For example, I had a "/codex-review" claude skill for ages that did exactly that. Technically you're right it wouldn't be switching, since you're right the two ideas are at different altitudes, but I think in practice it has the same impact: within one harness, you can delegate certain tasks to certain models or harnesses.