Imagine if you locked up your product and only slightly improved it every year and kept it locked up. No competitor, no innovation, just a stale product riding it out. I see no point in that. Competition keeps everyone on their toes. Makes you want to be the best and better than any other competitor out there.
From NVIDIA’s earnings report:
Commitments to purchase semiconductor components surged to $390 billion from $279 billion in the prior quarter, driven mainly by multiyear agreements to secure memory supply.
I think it’s okay for Jensen to take a break. He had a good long run and it’s fair to expect him to disappear for a while. Shouldn’t leaders be held to the same standards as everyone else? You don’t get a hall pass just because you run a successful company. Nvidia will survive this. It is resilient and it will be okay.
@MTSlive Maybe US companies like Anthropic and OpenAI should release their own open models. So we can discourage the use of “Chinese” open-weight models. They won’t do that, do they?
SITUATION DETECTED: The White House and Treasury are discussing a proposal that would require US companies to publicly disclose the use of Chinese open-weight AI models on their websites, aiming to discourage their use through public pressure, per Reuters.
Introducing Pirate Face
The first verifiable torrent index for AI - every Hugging Face open model, provably identical to the original and kept alive forever by the swarm.
https://t.co/6UpdzEuYTI
The permanence layer for AI. How it works 👇
What a day already! Jensen's apology letter, NVIDIA earnings, OpenAI launching a disastrous 6.2, Dario facing the congress, what else did I miss today?
Well now that we officially have a non preview Qwen 4 Max - the hype is dead. There is no significant improvement whatsoever. What a let down. I don’t know if the small denser models are any better.
@Supermicro We need a banana for scale. Everytime I see one of these online I have no perspective how big these are. These are massive irl. Supermicro but not super micro. lol
Qwen 4 Max Preview was underwhelming so far. It does good at times but it is no huge improvement from Qwen 3.8 Max or any other Qwen Models. There was so much hype to it and I hope the non preview version redeems itself. If not I will be looking forward to DeepSeek 4.2 I guess.
Super duper sick of scanning my fingerprint every other minute. Don't say "just get a YubiKey!" a secure exchange should run behind the scenes. i don't need to do a whole bunch of stuff to prove its me. We have a lot of smart machines out there and I believe we can get them to detect us without having to show my proof every minute.
Microsoft's Company portal is one of the worst apps I've seen recently. The fact that there are companies that still use it is mind boggling. Every few days it forgets to "trust" the device I work on and its like "Who the f are you??"
I re-register the device and it goes "oh yeah its you again" :|
With the cap, all 4 builds loaded with zero console errors: a landing page with a working pricing toggle, an original canvas game, an SVG dashboard with filters, and a physics toy holding 60 FPS with 150 bodies.
Two needed one "continue" call to finish.
Qwen4 Max Preview testing results:
I gave Qwen4 Max Preview a hands-on test today: 29 auto-graded tasks, a 42k-token needle search, a speed run, and 4 single-file web builds with no libraries or image URLs.
The screenshots are exactly what it shipped.
Biggest gotcha: with no thinking cap, 3 of 4 web builds spent the entire 32k output budget reasoning and returned zero HTML, after 12-14 min each.
Setting thinking_budget=8000 fixed it. Cap it.