Keep on to blur preview images; turn off to show them clearly
Chinese need to graduate out of this insecure mindset. "grooming is for f*ggots" is prison gay mentality. you don't have to compete with androgynous Korean twinks but nothing is more laughable than this Russia-style "a man must be only a little hotter than an ape" cope
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1
All selection tools are now working with depth-based constraints, making it much easier to clean a Splat in LichtFeld Studio.
Founder and CEO of https://t.co/5MjtfpwEU3 | Your guide to radiance fields | Host of the podcast @ViewDependent | FTP: 279 | discord: https://t.co/lrl64WGvlD
Google really outdid itself with Nano Banana Pro. Combining the image model with true intelligence opens up so many cool use cases. I took a picture of my daughter yesterday and made Gemini change her clothing to be dressed like a Zouave troop (French North African troops from the 1800s). She looked so cute in it that I wondered if I could somehow turn the picture into a pattern to actually make the garment. This is the kind of thing that was simply well beyond the capabilities of these models even 3 months ago. Now, it seems like they’ve already cracked it, And it will be even better in 6 months, and we’re probably only a year or two away from cheap custom clothing on demand that’s cut and sewn by robots according to AI-generated patterns (who is building this btw? I have a lot of weird old uniforms I want made for my kids…) We are living in the sartorial future.
This remains my favorite use case for AI by far.
the number of well-executed easter eggs in @stripe's BFCM city is just incredible * shoes outside of @cursor_ai * hover on ppl like @levelsio * Cheeky Pint bar * @tryramp ramp * @bannerbearHQ bear * actual @TBPN audio * live transaction volume what else?
Also looove to see many of my favourite indie hackers make it on here @dannypostma's Headhost Pro @JohnONolan's @Ghost @yongfook's Bannerbear @Erwin_AI's @lexboostai @levelsio as himself lol Years ago, most ppl thought nomads were living unserious lives. How things have changed
i think we have strong hints towards 4o being continued as base until o3 at least. but there are two points for which i do not think "no successful pretrain after 4o" is true. > several occasions on which i could see more recent additions in training knowledge as a user (midtrain only?) > there are many stronger base models available in the open now with more data as well that it does not make sense about oai specifically failing at the same size failing to scale RoI beyond 4.5 base size due to various reasons? yeah possible. failing to do anything better than 4o base at all, would not believe. that being said 4o being everywhere would also make sense in a scenario where there were certain internal or external evals related to user satisfaction where it outperformed everything and they needed to keep it as a merged base or a variant. but i wouldn't think it is because of incompetence in pretraining.
making models learn • eXperiments lab • memes and training lores
RT @ludwigABAP: S.A has nearly completely fallen off bc ppl cannot handle a bit of clout and fame and immediately get lost in the sauce
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1