OpenAI and Microsoft's partnership is ending... Thank you Browserbase for sponsoring! Check them out at: https://soydev.link/browserbase Source: https://x.com/OpenAINewsroom/status/2049228769135874295 Want to sponsor a video? Learn more here: https://soydev.link/sponsor-me Check out my Twitch, Twitter, Discord more at https://t3.gg S/O @Ph4seon3 for the awesome edit 🙏
ADVERTISEMENT
I always thought, before ChatGPt we all agreed that AGI was defined as Star Wars Droids or Irons Mans Jarvis was the benchmark.
Loved that one! By the way: Maybe you should remove the thumbnails where you're squinting, because you don't actually squint.
Hey if making noise and having to delete something is the ONLY way to get Microsoft to listen, we thank you for having the platform to do it with.
33:18 we got theo bullying MS before GTA 6😂
sounds like hoolie
At 34:38 is FP4 really sufficient for generating anything but AI slop???
You really are in a unique spot! I tend to think of you as a dev running a startup, but it's more than that. It's probably best to realign my view to something like Elon-Minus-Twenty-Years.
It's insane how Microsoft could quickly fix their API performance in a couple of days. I can only imagine on how many customers they lost out on because of this...
thank you some nice AI history
I'd take that 1M of Azure credits off of your hands if you don't want em, lol.
make a video about mimo 2.5 pro and deepseek V4
30:07 Theo's comment says the complete opposite of what he is yapping about in previous minute, good job Theo
damn really crazy how one dude makes an entire division in a trillion dollar corp improve. really fking weird to imagine tho you wonder what they are doing.
Obsidian checking in
It's hilarious that that last chart where Microsoft made improvement so fast is telling me even more to not trust them 😅
no video about deepseek v4?
Tbf I see speed issues occasionally with Sonnet on Bedrock.
Consider that they were secretly, purposefully throttling. Of course they would laugh until it becomes publicly noticeable.
I work at a non tech company building internal AI tools (healthcare). We literally only use Anthropic models because we already use AWS for everything so setting up extra networking work to use OpenAI models is too big of a pain compared to just using Sonnet or something
Chain of thought absolutely did exist as a concept before o1. It was invented at Google (as most things are) as a prompting method to improve accuracy by forcing models to reason (Jan 2022), then again Google were the first to train a model on CoT chains to create Flan-T5 (Oct 2022). Of course, there were more papers after that, but as usual, OpenAI was the first to scale it to the next level where it became relevant outside of research.