@mcuban Allow illiquid founders to pledge shares with a loan from the government to pay tax. The loan period is long but not infinite (e.g. 10 years). The loan is non-resource: at the end of the period, the loan is either paid back in cash, or the government assumes the shares.

Mathematics. I love this animation by Daniel Piker (@KangarooPhysics). Each dot follows a path, and takes 3.5 hrs to return to its starting point. (You might think the dots are jittering or sparkling, but on closer inspection they're walking like ants.) Used with permission.

This chart is the single best argument for why neoclouds are about to print money (Save this). SpaceX is generating between $30 million and $50 million in annualized revenue for every active megawatt of compute capacity, while pure play neoclouds like CoreWeave, Nebius and IREN sit in the $9.4 million to $10.4 million range. Traditional colocation players like Digital Realty and Equinix trail even further behind at $3.5 million to $4.4 million per MW. That gap matters because it shows exactly how much upside exists if neoclouds can push their revenue per MW closer to the top of that range and the mechanism that gets them there is simple: GPU rental pricing. GPU lease rates have been rising fast which is the opposite of what most people assume about a commoditized rental market. One year H100 contract rates jumped nearly 40%, from a low of $1.70 per GPU hour in October 2025 to $2.60 by now. This is essentially a self reinforcing cycle where tightening supply drives price increases and those price increases push neoclouds to lock in more hardware which tightens supply again. On demand pricing is even more extreme because every GPU model is essentially sold out on demand right now, with Blackwell generation B200 pricing running $4.99 to $18 per GPU hour depending on provider. Several neoclouds have already started raising published rates rather than cutting them, with Lambda moving from $2.99 to as high as $4.29 an hour and Verda climbing from $2.29 to $3.25. This pricing power flows directly into that revenue per MW chart, because every megawatt of power a neocloud controls becomes more valuable the higher GPU rental rates climb. Rising rental prices expand return on invested capital for deployed GPUs and extend the economic useful life of existing hardware, meaning neoclouds squeeze more cash flow out of the same physical footprint before needing to reinvest. That's the real bull case underneath the chart because power and megawatts are the scarce, fixed input, since Gartner expects power constraints to limit 40% of AI data centers by 2027, while GPU lead times already run 36 to 52 weeks. If a neocloud already has power secured and GPUs deployed, rising per GPU hour pricing translates almost directly into rising revenue per megawatt with minimal added capex and that's precisely why CoreWeave, Nebius, and peers sit so far above legacy colocation players on this chart. Colocation companies just rent out space and power but neoclouds capture the pricing upside of the actual compute running on top of it, and as GPU scarcity persists, that spread between neoclouds and traditional colocation should only keep widening. Bullish on Neoclouds, make sure to follow @MelvinInvests for more AI infrastructure insights and if you want to see exactly what I'm buying as an analyst at Milk Road Pro, you can check out the link below for more.

Announcing Discovery Loop! I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine learning, science, and engineering to accelerate discoveries and progress. The four of us have worked together for 14 to 30 years, and have helped build some of the world’s most used products, infrastructure and AI models, and we’re excited to turn our attention to this ambitious endeavor. ♾ Learn more at:

It took me a long time to build an intuition for why CoT works. My thinking was always.. if the model can predict it downstream of 10k thinking tokens, it should have been able to predict it from the outset too. My intuition now is: - During inference, the correct paths are indeed somewhere in the hidden states, represented purely as probabilities - However, in the process of sampling, we're forced to materialize just one path. This is destructive -- a 30% chance of ending up at the answer can become 0 if we sample the wrong token. - The constant backtracking reasoning models do protect against this. Every "wait" or "but" is another chance for a shot on target. - By the time models exhaust their reasoning budget, they've already seen a bunch of possible answers - And since these models are also generally better at verifying answers than generating them, the chances of choosing the correct path, conditioned on this prefix, are much higher than it was at the start.

the fact that founders of ai labs so regularly leave the companies they started, often within months is raising or not long thereafter before anything has really happened, should somehow register as a reward signal that: a) you should think really hard about your motivations to start a company b) doing so should be because there’s just no other vehicle for you to express your talent and conviction on bringing something new into the world and doing whatever it takes to win c) you have to sure about your choice of cofounders, come rain come shine otherwise, just don’t do it, and that’s fine!

Twitter really is its own universe... All I see here are "ChatGPT/OAI will rule everything!" and "no other model matters!". I think VCs internalize this. The real world is "I need to control my model's output and want to train/fine-tune my own." "I don't want to copywrited information in my model." "Smaller models are a better app experience." etc. If the real world speaks, one should listen...it's going to be a lot more complicated. Real apps will be built of bespoke models, APIs, DBs, and code.

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models.

Around 15 years ago, Kaheer Suleman and I started Maluuba with the vision of building Universal Turing Machines. Back then, we were one of the pioneer research labs building foundational modules of deep learning and Reinforcement learning with Turing award winners @Yoshua_Bengio and @RichardSSutton. There was no AI hype then, as there is now. Those were the good old days of doing pure scientific research unlike the n+1 research that we see today. Maluuba was later acquired by Microsoft and became @MSFTResearch Canada. Grounded in that same mission, we're finally ready to introduce @skyfallai to the world after a lot of experimentation in stealth over the last year. We're a frontier neo lab building the first Autonomous Enterprise by moving beyond the current LLM paradigm. For the last 5 years, the foundational model market has relied on a single paradigm: scaling laws for LLMs - more data, more compute, bigger models. However, the real world is messy and much more complicated. In order to achieve our long term vision, the next generation of AI models requires a different approach. Skyfall is solving the hardest open problems in frontier AI: long-horizon planning, data inefficiency, and brittle performance in dynamic real-world environments. To achieve the team's vision of a completely autonomous enterprise, the team is developing a next-generation frontier model (Enterprise World Models) using Continual Learning and World Modeling. Enterprise World Models can simulate the multi-layered consequences of strategic business actions. Our approach unlocks a new category in the foundation model market. To prove it, we're introducing Morpheus, a Continual Reinforcement Learning platform for AI researchers. I'm building this company with the people I trust the most: my longtime friend Kaheer Suleman (prev. Co-Founder of Maluuba) and my brother @omgiamgod (prev. YC founder). Sumit and Kaheer are the first principles thinkers I can trust to go to the end of the world with to achieve the mission impossible together. Together with a stellar team of 25 researchers and engineers, we're pushing a new frontier in AI forward. We unpacked our long term vision in today’s Forbes feature 🔗- read it to see what we’re building toward. Thank you so much Victor Dey for the interview. To achieve our goal of enterprise world models, we are soliciting bids to acquire small SaaS startups (up to $1M) and fully automate them. If you’re interested, submit your business here: Finally, thank you to our investors and advisors for believing in our vision since day one: @Fidelity , @sk121 (@touringcapital), @karam_n and @chrisarsenault (@inovia), @morgan_blumberg (@M13Company ), @stephpalmeri (@NextViewVC ), and @michaellitt and @mmccauley (@GarageCapital ), @jennydhe, @fchollet @NaveenGRao and so many others for supporting us in this journey.