Sarah Guo

it’s poetic that the answer “spend more test time compute” is so powerful most things respond well to human effort, too many apparent human limits are not “capability” limits. they’re budget limits (will, time, patience)

as mentioned in our Build interview — @satyanadella and @Microsoft betting on an ecosystem winning AI https://t.co/VT6VYtNdBV

California’s next set of leaders should take notes from Greg Abbott on how not to chase off a golden goose that can fly to space, create wealth, and build local industrial capacity from scratch

break out of the prison of the mind. and the earth. ad astra! congrats SpaceX friends

The class I loved best in school was architectural history. I am not religious, but visiting Sagrada Família was one of the most moving experiences of my life. I am awed by the generations who have built it, and the people and church that have supported it, and by its beauty.

new @NoPriorsPod with Priscilla Chan, Mark Zuckerburg and Alex Rives: - taking seriously the @biohub mission to cure and prevent all disease (soon!) - Model release of ESMFold2 and ESM Atlas (beating AlphaFold) - new biological knowledge from the models - ecosystem strategy https://t.co/oMItzKgvSV Full interview on YouTube or anywhere you get podcasts: https://t.co/ae6CycwnO9

C.S. Lewis is always a good re-read but especially at this time

The Untrainable

The mid-2026 investor's version of AI psychosis is a despair that nothing is investable, that we should put all our money into Anthropic and Nvidia and go home. I have never felt it. I have been sure the models are smarter than me for several sub-versions now, I’d be a happy buyer of Anthropic and Nvidia at the market price, and all my smartest friends are quite convinced that self-improvement is soon to work – and I still don't feel it. The despair isn't stupid. The logic runs: if the model keeps getting better at everything, then every company built on top of one is a thin wrapper waiting to be absorbed, and the only value that survives is …

cursor, lovable, cognition numbers all a big narrative violation. wasn’t everything in the path of agi labs (especially the #1 fight, coding agents) supposed to die, not accelerate

1/ thinking: so many default behaviors of models are based on the failures of internet data. secondary framings like rankings, summaries, takes, etc appear far more often in training than the primary sources they compress. 2/ so they’re the low-effort, “feels fluent” default, and the model emits them from memory instead of reading the source, unless user/agent forces it to retrieve. they don’t want to think critically! in novel ways! it is hard! 3/ folks broadly complain about models being “nerfed” but interesting to think about specific bad behaviors in models that I expect might get worse as companies “compute/cost-optimize” their general…

just watching the train of thought for any of the frontier class models is simultaneously thrilling and so disappointing. it fails by being generic and it knows it! it’s so repetitive! it cannot juggle enough ideas in its brain at once!

@ChrisPainterYup How exactly will atoms stop being an issue? Interconnection queues, power generation, the physical labor for standing up datacenters, your semis/memory supply chain bottleneck of choice, and the lack of “the internet of data” for robotics are solved by improved cognition?

@vixsheikh yes, he saw illegibility as a form of freedom

Scott in 1998: legibility is a central problem in statecraft. states often knew precious little about their subjects, and flattened them so they could count, tax, conscript, and control them mētis is the counter concept, the value of local expertise interesting in 2026 too https://t.co/6AJ77gZfT4 https://t.co/AI6wePSJeQ the author has passed, here’s the pdf

a “fast” takeoff with slow diffusion (gated by atoms and institutions and people’s discomfort with change) looks, from the economy’s vantage, less discontinuous

@ericabrescia I agree — I mean at more scale (tens of $millions+) when the growth number actually means something. Early growth is just super noise

there seems to be a “rule of 40” for AI businesses if you are growing >3X yoy investors seemingly accept negative to low GM and ignore questions about durability or, you can grow “fast” but not insanely, and you need a much better moat (models, enterprise) weak on both = tough implied here is a belief in the value of scale (as one rough swag, reach of $500M+ ARR to enterprises or consumers)

@Hershal0_0 @Microsoft @satyanadella @swyx @eladgil glad to be of service

@tpottery_ @howard nope? @johnolilly said this word to me 100y ago and I looked it up

@DamianBullish no. feel free to run statistical analyses on us screwed folks that have been using emdashes since they were taught in middle school — sad! But — true. @DamianBullish also, do you really need AI to write *two sentence fragments* on behalf of a friend who is hiring?

so fun to work with MacInnis, on an iconic company! storytellers — juicy job

@zeroXmusashi no, it’s really not. you can have conviction in a clear point of view that isn’t “let’s not reason, and instead assume doom/rapture”

Congrats team @chaidiscovery — a team that is not only training AI models to accelerate science, but rapidly deploying them with the leading pharma companies of the world We will see better therapies, faster because of Chai 🧬

@natolambert good man doing good work

@aaryendersingh @NoPriorsPod is the popsicle actually a bunch of peptides you inject

“Not AGI-pilled enough” is a perfect SV insult because it turns 𝑛𝑜𝑡 𝑡ℎ𝑖𝑛𝑘𝑖𝑛𝑔 into a status position. No need to have a timeline, strategy, theory of deployment, view on institutions, or second-order effects. Just stare at the eschaton harder than the next guy. begone ye shibboleth. turns out belief intensity is not actually a proxy for intellectual seriousness...

the non-tech public needs to be presented: - the benefits of AI (products, services and outcomes) - the ability to participate - the requirements (natsec, competitiveness) - mitigated concerns (jobs, energy cost) everyone has a real shot to participate in the new economy! the popular narrative (3-4 labs fight for the one ring, only they / their 10,000 followers will be uberrich, their models eat the rest of the economy, also it will destroy security, biosafety, and possibly kill us all) is 1. not true 2. not what everyone else wants

a lot of very successful companies look like a series of pivots (tacking?) in a general direction

1/ 🔥 @NoPriorsPod x @LatentSpacePod chat with @SatyaNadella at @Microsoft Build. He has the sharpest mental models of any public company CEO I've interviewed. $MSFT is at its heart still a tools company! Big focus on agentic coding, harness & AI evals. Takeaways:👇 https://t.co/iAAjHdCVnK 2/ Satya started with a reminder: to be a platform, more value has to be created on top than is captured first party. Microsoft wants a healthy ecosystem! Their goal: making any company, AI-native or legacy enterprise, a first-class participant with its own AI. 3/ Top quote: "Can everyone operate at the frontier with *their* frontier intelligence? T…

Congrats to team @TownAI! One of the first viscerally different AI experiences. Long horizon agents, in the places we do our work, founders you back every time 💥