Chamath Palihapitiya

agreed.

Deep Dive: The Frontier of Physical AI with Humanoid Robotics

The global economy has an impending problem. While AI is compounding its ability at a historic rate, an aging population and declining fertility rates are already causing labor shortages. These trends, combined with declining costs of robotics hardware, underpin a compelling case for humanoid robots and physical AI. According to Morgan Stanley, the humanoid robot market is set to exceed $5 trillion by 2050. Even in 2025, the larger robotics space saw $21 billion of VC capital invested. And with a steady increase in patent activity mentioning “humanoid” over the past few years, these machines are already walking onto factory floors. In N…

If…Not…When In my opinion, the current AI repricing of public companies is a process of asking “if” any future cash flows are safe. We used to debate “when” those cash flows would degrade - in 10 years? 20 years? Now the question is if these future revenues and RPO have ANY value. When you start to ask “if” questions about a company’s future, the safest thing to do is to re-rate their P/Es (lower), rev multiple (lower) and WACCs (higher) such that it embeds the holder with a massive margin is safety. This phase shift is very important and is what is currently happening. The downside of this repricing is that companies that are directly i…

A good reminder that idiotic tax polices like this can only be made up by naive theoreticians who have never had any responsibility in the real world nor understands how an economy works. Or the bitter poor. Those who own nothing and have no desire to work their way up and so would rather pull everyone back. Those are the only two kinds.

Stanford is home to some of the best software engineers, and our @8090_Factory HQ is right around the corner. We're actively recruiting interns at Stanford. Come work with us...

Good software doesn’t start in an IDE. It starts in a document or even a scratch pad where product thinkers describe and map their intentions and desires. In an AI future, the advantage is in writing clear requirements and engineering blueprints from these, where the intent becomes the structure for AI and agents to do the heavy lifting with minimal drift or guardrails. That layer shouldn’t live as a passive editor. It needs an agent collaborator that understands the product and helps turn ideas into executable plans. We’ve built that into Software Factory. This week’s upgrades include major improvements to the agent suggestion interface,…

Reels make you retarded. We should get our adversaries addicted to Reels. Hey, wait a minute! 🤔

Why?

OH: “The democrats really make themselves look either stupid or corrupt, with their disingenuity.”

Making A Machine That Builds Machines

Before the Factory, There Were Craftsmen In 1771, Richard Arkwright opened a cotton mill in Cromford, England, and the world quietly split in two. Before that moment, cloth was made by hand - one weaver, one loom, one room at a time. These skilled craftsmen were indispensable but they were also slow, expensive, and structurally impossible to scale. A merchant who wanted more output had only one option: hire more craftsmen and wait. I propose that software in 2026 looks almost exactly like that pre-industrial workshop. Need a product built? Hire engineers. Need it maintained? Hire more. Need to migrate off legacy systems? Budget two years…

China added more solar capacity in 2025 than America has installed in its entire history. That's the most important energy chart you'll see today. And 2025 was also the first year when small-scale distributed solar pulled in more investments than utility-scale solar farms globally. Considering that the U.S. has hundreds of GW stuck waiting for grid connections, the conditions are aligned to start putting solar + storage on every American home. My research team put together a Deep Dive on solar, if you want the full breakdown. Here’s the link: https://t.co/NppxPJDM1s

California sits at the intersection of two extremes. On one end is a population of extremely smart, capable Californians that have helped build the 4th largest economy in the world. At the other end sits the same extremely smart, capable population but who have turned their brains off and have elected, year after year for decades, a group of incompetent and corrupt politicians that have hijacked the governance of the state. In lockstep, the finances, pensions and livability of California erodes every day: the fire hydrants are empty, the drug addicts get free drugs, the alcoholics get free beer, the criminals get “get out of jail free” car…

It should be obvious to anyone that reads the list below that many of these companies will not survive as companies in a world of AI and Agents. These are services. Some more useful than others. But none of them control any capability or rails that couldn’t be negotiated/built (from open source or otherwise) by an agent for much cheaper. The alternatives will come with no capex, no OpEx, no rent seeking etc etc. The cost advantages of these things being rebuilt as primitives and repriced so it’s principally available to other agents and AI is pretty obvious.

In 2025, 25 data center projects were canceled due to community pushback. That’s up from just 6 in 2024 and 2 in 2023. The opposition is notably bipartisan, driven overwhelmingly by one thing: rising electricity prices for local residents. In Q2 2025 alone, 20 projects were blocked or delayed, putting $98B in potential investment at risk. The 2025 cancellations represented ~4.7 gigawatts of lost electricity capacity. Using OpenAI’s own estimate of revenue per gigawatt (~$10B revenue per gigawatt), those cancellations represent ~$50B in lost AI revenue in a single year. Applying a 20x earnings multiple and its $1 trillion in lost enterpris…

@Jason 🙏🏽🙏🏽

Facts.

Wow. As mentioned on @theallinpod a few weeks ago, there is lots of decode silicon coming…

Everything about cancer sucks. Some of the most difficult decisions in cancer are what to do and when to do it: biopsy, resection, passive monitoring etc. Today, clinicians make irreversible calls like 1) do a biopsy or wait, 2) treat or monitor or 3) remove an organ or monitor, using tests that cannot see the full biology of either the body or the cancer. As a result, doctors often overtreat the wrong patients while missing dangerous cancers in others until it’s too late. Five years ago, a sibling trio of Purdue grads cold emailed me from Indianapolis. Their thesis was simple: the science in cancer detection and treatment isn't the bottle…

Launching a YouTube channel to share business stories, lessons I've learned, and important topics I’m actively exploring… https://t.co/qWi07Rr4aD let me know what you think and what you'd like me to share on this channel think would be fun to do a video on my best and worst investments

Deep Dive: How Machines Are Becoming Better Investors Than Humans

Humans are not wired to win in today’s modern markets. For most of history, investing was about relationships, judgment, and timing. It was about who you knew, what you believed, and whether you had the conviction to act. That edge worked when information was scarce, and markets were slower. Today, public and private information are instantly priced in. As such, strategies that depend on reacting to news, reading filings, or “having an opinion” compete in a game where the half-life of new information is measured in milliseconds. No human, no matter how smart, experienced, or well-networked, can manually process that firehose of data in rea…

Another week, another Software Factory release: • Agent Framework UX Overhaul: Scroll control during streaming, real-time visible thinking tokens, and one-click navigation to retrieved sources • Structured Question & Answer Tool: Agents pose questions via a Q&A form (free-response and multiple-choice). Users answer inline, skip, or submit partial responses • Context Window and Compaction Display: Progress ring in the agent panel shows real-time token usage against the context limit. Compaction runs during streaming with a visual marker where compaction occurred • Reduced Token Overhead: Restructured system prompts and preprocessed context…

I agree. @bgurley has written something very special. I highly recommend.

Model builders overfit for evals that consumers have zero idea how to interpret. Libraries of important and controversial prompts with side by side comparisons of answers, like the below, are much more interesting and useful.

I have a simplified conjecture about AI progress. Your mileage may vary but here it is. f(i) = p × c × a where i=intelligence, p=power, c=compute, a=algorithms, multiplied together. This means neglecting any variable creates a ceiling and any breakthrough compounds across all three. One caveat is that today's advances are more accurately shaped by the entire history of prior investment, not just a discrete investment or breakthrough in time. In other words, the variables have memory over time so that must be factored in. In any event, the chart below is interesting because it shows the major tech companies operationalizing this function……

Here is a cold email I got from Russell Straub last night about 8090. I asked him if I could share it. Who is Russell Straub? He’s a version of many of us: entrepreneur, more than 25 years in business, a capable team, looking for a way for himself and his team to grow through AI. If 8090 can help businesses like Russell’s thrive through the AI transition, we can help many millions of employers and employees and it will have been time well spent. “Chamath! After hearing you talk about it for many months, I was intensely curious to see if 8090 could be the platform to enable our small dev team of seven to 10X themselves! I personally had …

Are you running a large-scale modernization program at your company? 8090 can help you move faster without sacrificing quality, with a customized version of our Software Factory - what we call a "Factory Line" - that can: • Reverse-engineer many codebases across multiple repos and auto-generate authoritative business and technical documentation on those systems • Forward-engineer into target architectures • Optimize code as it runs to enable you to trim down systems (by deleting functions no longer necessary) • Validate and certify that your new systems are correct If this sounds interesting, we'd love to talk: factory-lines@8090.ai

@Sam_Mindset This is roughly accurate.

Ingest Structure Learn (ISL) is the new ETL. It used to be the case that a company would try to license this kind of data as an “edge”. I’ve seen many companies in SV try to make this claim. That is no longer possible because it is no longer an edge. When data like this is widely available, many services will ingest it, clean it up then learn from it. Some will make incredible breakthroughs. Those that learn better aka are more intelligent will win. Those that simply do the mechanistic work of ingestion, no matter how complicated or cumbersome, won’t have anything of value to show for it so it’s not worth trying. ISL for the win.

Undeniable for anyone with a brain

Well, it turns out the billionaire tax proposal that is upending the California budget was never about healthcare, it was about retribution and envy…