482 karma · joined June 15, 2017
There's a number of US providers who also run it, if that is your preference.
Silicon Valley is just so disconnected from reality.
And to put that into policy gets rid of the scale for everyone. You can see it with abortion restrictions in various states. Instead of the doctor's expertise, the lawyers are the ones to decide.
Then again this is very much on point for the US. There are no experts other than lawyers. /s
Isn't this just because foreign students pay more than citizens? Isn't this just capitalism and the free market efficiently allocating resources?
Something about 'having the cake and eating it too'.
Of course I am exaggerating, but this is not a 1 dimensional problem.
Why can't they be as excited as people already invested into Datacenters? /s
The hard problems still remain.
Happy to see unsloth making it even easier for people like me to get going with fine tuning. Not that I am unable to I'm just lazy.
Fine tuning with a UI is definitely targeted towards hobbyists. Sadly I'll have to wait for AMD ROCm support.
Although I have been a skeptic of MCPs, it has been an immense help with agents. I do not have an alternative at the moment.
Am so much more excited about tiny models gaining real intelligence. Just today I have been running Qwen3.5 0.8B model on images and am pleasantly surprised by how good it is compared to even 4B and 8B models from a few months ago.
With a single Tilt file combined with a docker compose file, almost all of the infrastructure you need is configured on a local machine. It also supports running kubernetes (most of the docs are around this), but you do not necessarily need to it it.My goto when I have more then 2 docker containers/services I want to keep changing code for. Some teams I work with usually have 20 such containers for local dev.
And yes, you can even nest Tilt files and even write normal python if you want to mix things up.