@ryanlightbourn@MattisRedacted Leaving a comment to look back in 5 years to see if this held up. I think we'll be beyond what is possible in a proper film today.
Yes but who would have paid all these people you're mentioning? Realistically most of the stuff that is done with AI by small teams or individuals would just have never existed because of a lack of funding, connections and the C-Suite not wanting the risk. It's literally the biggest unlock for artists that ever existed and yet people try to gatekeep it's use, as if that would make it go away.
Go do cool stuff. It's here to stay if you use it or not and the models are finally good enough that you can really decide on every detail.
@rezoundous Look if I can get through a full week of work then I'm happy. If I can spend some extra tokens on side projects even better. Currently I can.
trying some new variations of the model picker for the most used... There's too much white space on the new thread view and the picker too much of a hassle for it to not be used.
It's not about Teams. It's just that the EU and every company in it relies heavily on the US for everything software. Not a good spot. We outsourced most of our software + dev knowledge to the US. Just as the world currently does with manufacturing in SEA and China. It's not healthy.
Creating something worse than MS Teams is a real achievement. I do think it's good for the EU to build their own stack... it'd be nicer if we encouraged and supported startups more and making it easier to invest in them, so they could take this on instead of some government entity.
It's not purely about intelligence. Reliability is way more important. A lot of knowledge can be collected and saved, procedures can be figured out and saved. Day to day GLM 5.3 Flash is very capable.
You can't one shot a new graphics engine like you can with Opus 5.5 but usually that's not what you need every day. Imo a local stack + frontier paid per token is the sweet spot soon. There's not that much left that I need in my business and life that can't be solved in the background by something like GLM.
@MatthewBerman Some people are just inherently evil and just don't act on it too much because of consequences for themselves. "Some" is probably understating it.
@LLMJunky My biggest worry is legislation, especially in the EU since they always need to fiddle around in stuff they barely have understanding of. I'll feel more at ease once I got my local stack at least as a backup option.
@dkundel@camsoft2000 I'm already very annoyed by the delay until voice starts in Codex (if it works). And this seems to be like a good 6 seconds of wait? Why? Isn't it the same tech behind it? Also, EU when?
@Da7_Tech GLM 5.3 Flash is the first open weight that actually can run decently well for a long time and get real work done. Looking forward to 5.4 or whatever they'll bring next.
Still far away from Astra, Fable and Opus though.