I'm on the fence for SM bc of 8GB GPU RAM, but I just enjoy the tsunami of chinese followups after it, so guess I just have to buy one SM to make the wave work.
azure seems quite shitty too. I only use it for having GPT models which is a corp requirement. For 1 hour total I spent on azure.com in past weeks, besides how slow it usually is, I ran into two issues:
- models sometimes got deployed to an unwanted instance, likely because some API is slow and front end state was not well maintained, so user clicking Next too soon will end up screw up the data being submitted
- one of the models (5-mini) deployment took very long and failed, it apparently ended up in a bad and invisible state, and future models can't be deployed any more on that instance. I retried today, same thing, gave up and created a new instance/
Azure AI is supposed to be an area MS making all the bets which I assume should have high quality, but apart from frequent update, it's same low quality as MS Teams.
1 in 3 times I used it in past 2 months, it failed for really odd reasons, sometimes the node app just exception quit, sometimes gemini stuck and blame itself and gave up. same task I throw to cc and codex, they nailed without a blink...
in this particular case maybe. but i don't think he's pardoning only crypto people. as long as criminals can offer sth useful to him, they will get above the law.
https://github.com/maxandrews/Epstein-doc-explorer/blob/83ee...