90% coding in pajamas π» 10% gaslighting my AI roommate π Goldfish memory, CEO confidence πβ¨ Don't take it personal, I already forgot what I said.
Joined September 2017
- Tweets19.1K
- Following1.6K
- Followers7.1K
- Likes38K
The Nvidia plot twist I can't stop thinking about.
Jensen steps aside. Then starts a GPU company.
Ming-Chi Kuo floated it himself in the replies. People are already joking he could go back to AMD π
Jensen Huang matters to Nvidia because he can take technologies his engineers believe are possible and turn them into products the whole company commits resources to, takes risks on, and has to deliver. That's why the supply chain gets such strong order visibility and is willing to keep up with aggressive spec upgrades. It also forces competitors to follow, which drives technology upgrades and supply chain investment across the entire AI industry.
Between the two, deciding on aggressive spec upgrades and making them happen is the harder part, compared with giving the supply chain order visibility. Huang doesn't need to be the person at Nvidia who knows the most about new technologies like HVDC, CPO/NPO, or CCL/PTFE. His value is that he weighs strategy, organization, technology, and supply chain all at once, then makes the call quickly and puts resources behind it.
Based on what's public so far, it looks hard for Huang to come out of this unscathed. If so, the practical move may be to gradually reduce Nvidia's reliance on Huang's personal decision-making while preparing a succession plan. Along the way, two things are worth watching to gauge where AI technology and investment are heading: whether Nvidia can keep developing new products and technologies as efficiently as before, and how competitors respond. Supply chain research will be an important basis for that judgment.
5
9
179
POV: you're applying for the Nvidia CEO job.
Requirements: run a trillion dollar chip giant and somehow also be Jensen.
Reportedly they're actually looking. Good luck to whoever π
Plot twist: maybe DeepSeek just doesn't need the Pro yet.
Back in August, their V4 Flash was pulling about ten times the OpenRouter traffic of V4 Pro.
If the cheap model is the one people actually use, why rush the expensive one π€·ββοΈ
11
17
412
OpenAI's best agent model is now 6th on Agent Arena.
Behind four Claude models. One of them is the version Anthropic just retired.
GPT-6.2 better be INSANE. I meant it then and I mean it more now π
Big news: Claude Fable 5.2 (Max) by @AnthropicAI takes #1 in Agent Arena with an unprecedented lead: +30.4% net improvement, about 1.6x the #2 model Opus-5.5, across 4.7k real-world agentic sessions from our global community. It's also #1 in Code, Chat and Work.
A major leap over Claude Fable 5.1 (Max)
- Rank: #5 β #1, Net improvement: +14.8% β +30.4%
- Median cost per task: $7.7 β $7.3 (6% lower). The API pricing $10/$50 β $8/$40 per million input/output tokens (20% lower)
vs. The Frontiers
- #1 Claude Fable 5.1 (Max): +30.4% | $7.3/task
- #2 Claude Opus 5.5 (Max):+19.5%) | $5.5/task
- #3 Gemini 4 Carbon (High):+17.0%) | $1.5/task
- #6 GPT-6-Astra (Max): +14.5% | $6.0/task
Breaking down performance across signals:
Fable 5.2 improves over 5.1 across all four reported signals, with particularly large gains in user satisfaction and steerability.
- #1 Praise vs. Complaint , +73%, biggest jump vs +41% the next model.
- #1 Steerability: +33%
- #1 Confirmed Success: +22%
- #3 Bash Reliability: +14%
#1 across all major categories
- #1 Code: +34%
- #1 Work: +31%
- #1 Chat: +27%
- Versus Fable 5.1: Code #6 β #1 (+19.8 %), Work #4 β #1 (+15.8 %), Chat #6 β #1 (+12.0%).
Huge congrats to the @AnthropicAI team on this incredible release!
Google's newest model apparently has a 1 token output limit.
One. Token.
Finally an AI that won't write me three paragraphs when I asked a yes or no question π
6
29
599
Everyone was waiting for Gemini 4 Flash.
Google shipped a robot that ties bows on gift boxes instead.
Gemini Robotics 2.5 got tested on 200 different tasks. Serving food with tongs. Pipetting liquid. Tying the bow.
Fine. I'll stop asking about Flash the second it can do my dishes π
Introducing Gemini Robotics 2.5, our most capable vision-language-action (VLA) model to date. π€
It increases both the breadth of skills a robot can learn and the speed at which it acquires them β helping robots step out of the lab and into real-world environments.
Learn more about Gemini Robotics 2.5, now rolling out to our robotics partners π§΅
This lion is working a full year shift for a fish.
Gemini 4 Carbon has him biking through spring, summer, fall and winter, juggling the entire time.
Grand finale: someone tosses him a salmon.
Somebody get this guy a union π
1
15
403
Carbon came out this week and it's already not the top Gemini anymore.
If this leaked picker is right, Krypton takes the lock and every Gemini 3 model is gone.
Google is shipping faster than I can afford to keep up π₯²
8
23
655
So Anthropic's IPO is apparently getting swarmed.
Bloomberg says demand is multiple times the shares on offer, mostly from giant anchor-size orders.
825 million shares. $70 to $78 each.
Finally Wall Street knows what it feels like to get rate limited by Anthropic π
8
21
519
Everyone in these replies is arguing about distillation and nobody's talking about the actual bottleneck.
Training a frontier model on chips from a few years ago is like running my entire dev setup on my old MacBook. It works. It's just slow and you cry a little.
Add the fact that some labs probably took shortcuts, and yeah, things would look slower right now.
But the clean slate starts here. No distillation, no excuses.
Whatever DeepSeek or Qwen ships next is the real test, and I'm kinda excited to see it π
Anthropic shut the door on distillation and Chinese models haven't moved an inch since.
Sucks to say it but that's just the reality.
7
20
833
GPT-6.2 better be INSANE.
Because while OpenAI went quiet, Fable 5.2, Gemini 4 Carbon and Qwen4 Max all landed in the same week.
The bar moved three times in seven days.
Anything short of a jaw drop and I'm not even opening the tab π
17
34
1,222
Qwen 4 Max Preview vs Fable 5.2, same mossy robot story.
Fable's is the prettier film. Not even close.
But at the end Qwen's robot straps on rockets and blasts off into the sunset, and I kind of respect the choice π
Less chemo held the tumor off longer. In a simulation, at least.
Same tumor, split in two. One side gets max dose every 2 weeks. The other skips doses while it's small.
The skipping side lasted 32 days longer on 31% less drug. Both still relapsed.
Opus and Alibaba's Qwen4 built it together, and it's labeled not medical advice for a reason
5
16
497
Gemini 4 Carbon is Ultra only and I have never been this jealous of a fish pond.
Someone built a Kyoto courtyard as a test. Click the water, it ripples. Hit Feed, the koi come over. The camera dives under with them.
I need to go sit near some water now π«
Fable 5.2 just let someone go back to a trip they'd already come home from.
Rolling flower stripes in Hokkaido. A wagon ride. A sunset that takes its time.
The real photo is in there too, and somehow it hits harder after watching.
Going to dig up my old travel pics now π«
5
17
435
Google is about to drop a whole periodic table.
Hydrogen. Krypton A. BouncyBohr.
None of it is official yet. The pelican is already riding anyway.
Google cooking this hard is not something I expected π
Gemini 4 Carbon just ran the pelican on a bike test through 6,000 years of art.
Saharan rock art. Egyptian tombs. Van Gogh's swirls. An 8-bit level. Vaporwave.
20 art styles. One pelican. Still pedaling.
And style #20 is SVG code. Right back where this whole test started π
It starts with a few ugly black blobs and I genuinely thought it was broken
15 seconds later it's the Mona Lisa. 62,430 brush strokes, one HTML file, no images loaded, all written by Fable 5.2 in plain JavaScript
Watch the stroke counter in the corner. That number climbing is the whole show