Pinned Tweet
Gemini 4 Hydrogen is 7th on the Intelligence Index at 53.9. At a $3 blended price it's also the cheapest and best-value model in the top seven, with 18 points per dollar against Carbon's 15. Google's 57% coding and 2.5× speed claims aren't verified yet.
1,191
For those of you who don't know, my other side project is an app called NeoTorino.
Ever wondered what an asteroid would do if it hit land, the ocean, or blew up in the sky? NeoTorino tracks the real ones passing Earth, using live NASA data, and shows what each could do: how big, how fast, and how much damage.
Coming soon ... or as soon as I can fit it round my AI projects.
5
2
499
AI has cracked the visuals. The cinematic look and the sense of scale shown here are on a par with films like Dune 3 that spend hundreds of millions on the same kind of shots. Sound and acting still lag, though acting is improving fast. Within 5 years, production will look radically different.
499
Bloomberg's "circular AI deals" chart doesn't present the deals well, which is why so many people are misreading it. Arrows running both ways make it look like the same money going round in a loop. It isn't. The labs' spending commitments are paid from revenue that comes from outside the circle: businesses and consumers paying for Claude, ChatGPT and API access. Anthropic alone hit an $81bn annual run rate in October.
The chart also draws every link the same size, so a $1bn stake looks as big as a $300bn cloud contract. Here it is redrawn to scale. Not a scam, but a leveraged bet that AI demand keeps compounding. Watch the labs' revenue against their compute bills, starting with Nvidia's results on 17 Nov and Anthropic's first quarter as a public company.
1
1
452
Sources: Anthropic S-1/A (SEC, 10 Nov), Nvidia 8-K (17 Aug) for the $105bn Ohio lease guarantee, OpenAI and Amazon releases (27 Feb), WSJ via DCD for Oracle's $300bn, plus Bloomberg, TechCrunch and Yahoo Finance. Market caps are at the 12 Nov close. Commitments are minimums or caps, not cash paid.
346
30 years ago I was a student volunteer at SIGGRAPH in LA when I shared a hotel lift with Jensen Huang, who'd only just co-founded NVIDIA. We'd both missed the bus to the convention centre, so he called a cab and insisted I ride with him. That small kindness has stuck with me ever since.
Meet Gemini 4 Hydrogen.
It brings cutting-edge intelligence to everyone, with speed and efficiency for everyday coding and agentic work.
Across 4 thinking levels, it stands on its own or teams up with Carbon and Argon — 2.5X faster than 3.8 Flash, outpacing competing models in its class.
Gemini 4 Hydrogen, the Flash tier. Business Insider reported on 11 November that Google staff were testing it internally, and today some users are seeing it in the Gemini app. Google hasn’t announced it or published pricing yet.
304
The latest from the SI frontier: Anthropic’s model timeline now includes the still-unreleased Fable 5.5.
214
To be honest, I see this as @JensenHuang being human. We all make bad decisions, and what we learn from them can push us to do better.
I’m disappointed that @nvidia let a personal matter spill into a business decision. This feels like Apple ousting Steve Jobs. Jensen’s contribution went well beyond the GPUs. He could explain what Nvidia was building, why it mattered, and where the company was headed with conviction and authority.
I see two possible outcomes. Within two years, Nvidia could be asking Jensen to come back. Or he could start a rival company, draw some of Nvidia’s best talent, and eventually overtake it.
Either way, I think Nvidia has made a serious mistake. The fallout could reach well beyond the company and slow America’s push to lead in superintelligence.
Jensen, I wish you all the best. Keep doing what you do. You’re one of the greats.
248
F U Z Z Y G O A T ™ retweeted
NVIDIA announces CEO transition. Read the release: nvidianews.nvidia.com/news/n…
Here’s my snapshot of Google DeepMind’s model lineup. Orange marks today’s state of the art, and purple shows what I’m watching next.
Gemini 4 Carbon became generally available on 10 November and is now Google’s strongest model. It sits alongside Gemini 4 Argon rather than replacing it, and both are $2 / $10 until 30 November, when the price doubles.
Gemini 4 Flash is still a rumour, based on one unconfirmed sighting, and Google has given no date for Omni image and audio output.
With @AnthropicAI shipping Fable 5.2 and @OpenAI expected to answer with GPT‑6.2, Google has some catching up to do.
272
Top AI models this week, Fable 5.2 leads overall, and Opus 5.5 still wins coding. For the best value, Gemini 4 Carbon (general) and Sonnet 5.5 (coding) come close to the leaders at a fraction of the price. What to watch: OpenAI's next flagship and Grok 5 from SpaceXSI.
612
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges.
They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
The rise of AI is forcing us to confront an uncomfortable question: how much of what we think and believe is actually our own?
When we interact with AI, we often question whether its responses are genuinely meaningful, grounded in reliable information, or simply an amalgamation of artificial slop.
But what happens when we turn that same scrutiny on ourselves? How much of what we think, believe and do is genuinely our own, and how much has been shaped by external forces such as governments, corporations, advertising and mainstream media?
If we question the authenticity of an AI's reasoning because it has been shaped by external information, shouldn't we apply the same scrutiny to our own?
It's a deeply uncomfortable question, one that challenges the very nature of independent thought, free will and what we believe to be true.
77
F U Z Z Y G O A T ™ retweeted
Releasing the model weights of Kimi K3.1 Pro: huggingface.co/moonshotai/Ki…
We'll keep advancing open intelligence.






















