https://bannedbyanthropic.com/
Stumbled across this today and I have to admit the capriciousness of it put me on my heels.
I guess once you get big enough you can stop having to explain yourself or deal with customer service / resolution.
You must log in or register to comment.
Interesting: Sonnet 4.6’s performance has dropped significantly today and the status page is showing corresponding service issues. Probably the 4.7 rollout?
Random degradation in performance is one of the other annoyance of cloud based inference / another reason to roll your own.
The problem is that this is one of the few use cases where I can afford the (dangerous and unreliable) cloud service but I’m far away from being able to do so self hosted.
I’m actually using LLMs quote a lot to counter some of my brains more weird
bugsworking-as-intended features. But this is way beyond what I could locally do.For anyone in a similar shoe: you can mix local and remote quite well though, using local for everything that works out and remote for everything else. I.e speech to text and email analysis is done on my local server while bigger sets are done on the CLI with remote providers.
1·34 minutes agothat’s what I thought too, but apparently there are laptops these days that are incredibly powerful for AI. anything with the newish AMD Ryzen AI Max+ 395 for example. has 128gb of unified memory, so it can run very large models quite easily. it’s in the Asus ROG Flow Z13, which I’m very seriously considering getting. it’s been out about a year, so there are a lot of reviews out that have tested the performance with AI.
I think you’d be surprised at how common that use case is. Sentiment analysis, advice, help and general chat.
Technical coding stuff makes up maybe 5%.
Split as needed.
You would think that the premier cloud-based service would have that on lock. Their own internal polling shows the trend.
https://www.anthropic.com/research/economic-index-march-2026-report
What I think will end up happening is that the online providers will crater their good will as one stop shops and trend towards business use cases / API. Maybe that’s just me being a cynic.
No I think that’s the realistic way forward sans real AGI.
If you mean “keep as much local as you can, route to cloud for heavy tasks?”, then we probably agree.
IMHO there stuff that just out of reach for most consumer grade local rigs.
Even the bigger open weight models are now in the 1T range. The new Kimi 2.6 can technically be installed at home…if your home happened to have a $300K server.
I think maybe the real benefit of these large, competitive, open weight models is that they stop the established players (eg: OAI, Anthropic) from just cranking the costs upwards to far / running roughshod. If you can pay $10 a month to access something that is very nearly as good as SOTA, then…why would you pay $100 - $300?
