CEO @itsPocketWorlds @Highriseapp @Everskies | Empowering Creators, Game Devs and World Builders | Angel Investor
Austin, TX
Joined September 2007
- Tweets1.1K
- Following633
- Followers3.6K
- Likes9.4K
Anton retweeted
>be me
>discover effective altruism
>apparently normal charity is inefficient
>why donate to random sad thing when spreadsheet can tell you optimal sad thing
>fair enough
>buy mosquito nets
>save lives
>numbers look good
>feel powerful
>couple years later
>someone asks an innocent question
>why only count people alive today
>huh
>future people matter too
>obviously
>my grandchildren shouldn't matter less just because they haven't spawned yet
>reasonable.jpg
>keep following logic
>what about their grandchildren
>also yes
>what about people in 500 years
>sure
>5000 years
>why not
>500 million years
>starting to get weird but morality is morality
>open calculator
>humanity could survive for an astronomically long time
>could colonize galaxy
>could have trillions upon trillions of descendants
>maybe digital people too
>maybe simulated civilizations
>maybe dyson spheres full of happy uploaded minds
>calculator starts smoking
>realize currently living humans are rounding error
>8 billion people suddenly looking extremely beta
>future contains potentially 10^something people
>can't even fit beneficiaries in google sheets
>new moral priority unlocked
>protect the long-term future
>stop thinking in units of "people helped"
>start thinking in "fraction of cosmic endowment preserved"
>malaria?
>terrible
>but only kills existing humans
>AI extinction could delete the entire light cone
>nuclear war could permanently derail civilization
>bad institutions could lock in terrible values for ten million years
>someone invents wrong constitution in 2140
>quadrillions suffer
>better fund governance workshop now
>friend says maybe we should improve hospitals
>explain opportunity cost
>friend says hospitals are full of actual sick people
>explain scope sensitivity
>friend stops inviting me to dinner
>need to decide what to fund
>easy
>expected value
>suppose project has one in a million chance of preventing extinction
>sounds tiny
>but extinction destroys 10^50 future lives
>multiply
>mother of god
>$10 million project has expected value of several galaxies
>charity evaluation complete
>someone asks where the one-in-a-million number came from
>expert judgement
>which expert
>us
>how calibrated
>extremely thoughtfully
>reduce estimate to one in ten million to be conservative
>still beats curing cancer by 38 orders of magnitude
>epistemic robustness achieved
>someone says maybe project doesn't work
>assign 20% chance
>still astronomical
>maybe project makes problem worse
>assign 5% chance
>still astronomical
>why 5
>because 30 felt pessimistic
>publish 46-page report
>contains seventeen sensitivity analyses
>every sensitivity analysis begins after assuming intervention has positive sign
>critic says you're multiplying enormous hypothetical stakes by extremely uncertain probabilities
>yes
>that's literally why it's important
>critic says the uncertainty might be structural rather than numerical
>make probability smaller
>critic says no, I mean maybe your model is wrong
>make probability smaller again
>critic begins rubbing temples
>discover AI safety
>perfect longtermist cause
>AI might kill everyone
>or create utopia
>or seize galaxy
>or tile universe with paperclips
>or create billions of conscious software minds
>finally a problem with numbers big enough for me
>start AI safety nonprofit
>mission: prevent dangerous AI
>hire smartest people available
>smartest people immediately start building better AI to understand dangerous AI
>interesting
>we must understand capabilities to understand safety
>we must scale models to study alignment
>we must race ahead so less responsible actors don't get there first
>we must deploy systems to learn how deployment can go wrong
>we must build the thing quickly because building the thing quickly is dangerous
>outsider asks why the people most worried about AI apocalypse all work at AI companies
>complicated field
>company releases stronger model
>very concerned
>company begins training even stronger model
>extremely concerned
>company raises $14 billion
>concern reaches unprecedented levels
>need to influence government
>future is at stake
>normal democratic process too slow
>politicians don't understand exponential curves
>public doesn't understand x-risk
>experts must guide them
>who counts as expert
>people who understand x-risk
>who understands x-risk
>our friends
>someone objects that this seems politically convenient
>explain we're representing future generations
>future generations unavailable for comment
>develop concept of value lock-in
>terrifying possibility that one ideology controls civilization forever
>therefore extremely important that civilization adopts correct values before lock-in
>whose values
>let's circle back
>begin with impartial morality
>end with small group of people deciding what quadrillions of hypothetical beings would want
>beautiful arc
>meanwhile actual humans keep doing annoying things
>voting wrong
>having parochial attachments
>loving family more than strangers
>caring about local community
>getting upset when told their suffering is cosmically negligible
>evolutionary biases everywhere
>explain that moral intuition cannot be trusted
>except intuition that future digital people count
>and intuition that extinction is uniquely bad
>and intuition that our probability estimates are sane
>and intuition that our institutional choices improve the future
>those intuitions survived peer review
>someone donates $5k to local homeless shelter
>inefficient
>could have funded 0.0000000000003% of an AI governance researcher
>think of all the simulated people you just killed
>okay maybe don't phrase it that way publicly
>PR team says "future generations deserve a voice"
>much better
>journalist asks what longtermism means
>say "future people matter"
>everyone agrees
>great
>journalist asks what follows from that
>well technically we should redirect enormous resources toward low-probability interventions affecting astronomical futures
>journalist raises eyebrow
>return to "future people matter"
>motte has entered the chat
>critic: of course future people matter
>me: glad we agree
>critic: I don't agree that your institute knows how to help them
>me: why do you hate our grandchildren
>eventually notice uncomfortable implication
>if future value dominates everything
>then helping people today mostly matters through effects on future
>education matters because future institutions
>health matters because future productivity
>democracy matters because future trajectory
>human beings slowly become instrumental variables in their own moral philosophy
>see starving child
>feel compassion
>check spreadsheet
>child's direct welfare contribution negligible
>but perhaps childhood nutrition improves national institutional quality
>compassion restored
>tell myself this is impartial altruism
>one day assistant asks obvious question
>"how do you know your intervention actually improves the far future?"
>silence
>open spreadsheet
>increase column width
>add confidence interval
>assistant asks again
>"no, I mean how do you know the sign is positive?"
>stare into cosmic light cone
>10^50 people staring back
>none of them exist
>none of them can tell me
>none of them can falsify my assumptions
>realize I have invented the perfect constituency
>infinitely important
>completely silent
>and always represented by me
Anton retweeted
I got access to Jev earlier today (thank you @hackgoofer). I have run ~5,000 requests so far, (which cost me around $2!), across classification, model routing, intent, steering, and many other things.
tl;dr, Jev enables a new intelligent decision-making primitive, separate from deterministic code and LLM calls. This allows a class of decision-making that was neither suited to dumb, unintelligent code, nor to slow, expensive LLMs.
It is super fast and cheap, and I think I will likely end up making a few Jev calls to every LLM call I make in my product. I think probably any company using LLM requests today can probably add a Jev call pre and/or post LLM calls to quite literally make their product much better for free, and have better tool calling behavior in many cases.
I happened to have a personal benchmark for this as I’d been working on a ton of proactivity and classification tasks. I have been using the deepseek flash and more recently gpt 5.6 luna family of models as reasonably smart classifiers with low latency. Think questions like:
- Did this conversation output contradict something they’ve mentioned before?
- Should we send a followup message to this user based on our rules?
- It’s been a few seconds of silence. Should we proactively send a message?
In the past, I’ve been forced to write a bunch of what I call decision chains, mostly because an LLM classification call is very expensive in TIME (avg 4s), and less importantly can cost quite a bit if run on every message.
Imagine a normal chat app. If you added 4s to every response to figure out if the response is good before sending it out, that ends up being pretty bad. So instead, I usually have to write some code that is a crude heuristic that runs quickly and decides whether to run the classifier. Obviously this sucks because you call the classifier many times that you don’t want to, which makes your p95 bad, and you also miss cases with the heuristic, and you also have to manage all of these weird chains.
With Jev, it’s cheap enough, and fast enough (p50 ~150ms, p95 ~350ms in my testing!) that you can easily run it every turn. Heck you can reasonably run it before generation AND post generation, for any application that isn’t realtime voice, and still feel snappy.
But this is just one use case. Think: smarter model routing, better context packing, better responses, better observability for intent/tags/safety, smarter retries and so much more. By simply thinking about the inputs and outcomes you want to enforce, you can use Jev to supercharge most model calls and reduce bad user outcomes. The more “quirks” a model has, the more valuable it ends up being.
It’s a bit weird and unintuitive using Jev. Generally, you want to decrease the # of questions you ask a classifier, or it makes more mistakes. In fact, you might want to ask your questions kind of in a compound way, because the reasoning happens in a shared scratchpad of sorts. Adding questions muddies the scratchpad and makes it take longer.
With Jev, you feel incentivized to go the other way, to formulate your query as a set of independent questions. It doesn’t feel like adding more questions decreases your performance on others.
You can go a bit deeper to improve tool calls. Many model tools are things like turning on settings, or other things. You can easily improve models that are not very good at tool calling with Jev, by simply figuring out when to run them. You can do a pre-LLM call to figure out when to unfurl different tool definitions, in order to make your main LLM run better, you could run a background task with Jev + another LLM to reduce tool and context burden on your main LLM, and free it to be responsive.
I’ve only scratched the surface of my testing but very excited!
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI?
I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev
• 20-200x faster
• 40-400x cheaper (w/ output tokens free)
• Frontier composable intelligence optimized for decisions
AFAICT the shortest path to AI-based economic revolution
Narrative violation
Since the launch of ChatGPT, the Philippines offshore industry (8% of country’s GDP) has actually grown much bigger.
Employment in IT and business outsourcing is up by +20% to 1.9m workers and industry revenue is up +30% to $42B.
“AI is helping offshore workers in the Philippines do new, more complex jobs. They are picking up work training AI models or supervising AI agents. Hospitals in America are increasingly outsourcing the checking of insurance eligibility and filing of medical records to Filipinos packing AI tools. Mr Gallimore says more “high-value” work, such as accountancy or engineering, is going offshore in part because “AI is a leveller. You can now teach someone complex stuff quickly” and still get it done more cheaply than in America or Europe.”
***
More here: economist.com/asia/2026/08/0…
Anton retweeted
Someone asked me recently why I’ve become interested in aesthetics after having spent most of my life more interested in STEM-adjacent topics. I hadn’t really considered the question consciously before, but I’m certainly thinking about aesthetics more than I used to. I think it’s a confluence of things:
• Many things today are ugly and far uglier than they used to be or need to be. Once you see this, it’s kinda hard to stop perceiving it. (Early twentieth century phone boxes versus modern phone boxes; old water fountains versus new water fountains; etc.) As someone with a naively meliorist assumption that most things should be getting better rather than worse, it’s all a bit perplexing: why did we stop doing things nicely? Is it a choice? Was there a malevolent spell cast upon us? This vein led me to think more about modernism and why much of art became more intentionally "challenging", grotesque, opposed to prettiness, rebarbative, dissonant, etc. Can or should anything be done about this? Is this just how things ought to be?
• Relatedly, much of modernism involved a kind of explicit repudiation of cultural continuity and represented a schism with prior practices. This is maybe most evident in American architecture, where the International Style exhibition in 1932 initiated the displacement of a rich tapestry of prior styles. This cultural break seems important and interesting to me, and I suspect that the rejection had important consequences outside of the aesthetic domain. Samuel Hughes has been exploring this question in his recent writing at @WorksInProgMag; @RuxandraTeslo is also pulling on this thread. Elaine Scarry wrote about how beauty inspires creation. If so, the inverse may also be true: ugliness inhibits it.
• It’s clearly the case that changes in the aesthetic domain can at least inspire progress in other places. Petrarch helped set some of the preconditions for the Renaissance which in turn fostered the scientific revolution and Enlightenment. Things like World Fairs (the 1851 Fair at the Crystal Palace recorded 6 million admissions when the population was 21 million) reflect the popular interdependency that used to exist between aesthetics and material development.
• @tedgioia and others have written about stuck culture and how so many domains seem to have ceased to straightforwardly advance in the way that they did up until the nineties or thereabouts. This is obviously peculiar and interesting. What changed, and what does it mean? Is it about the internet and fragmentation? Is it about a loss of supply? Is it just about having reached the zenith of various mediums?
• I’m generally interested in markets and the dynamics of creation. In aesthetics broadly, I find the reflexivity between supply- and demand-side factors to be very thought-provoking. There’s a natural desire to view satisfaction of individual preferences as the yardstick to measure market success, but things get interesting and even a bit unsettling when we start to think about how the supply might start to shape the demand. I often think about this in the context of food. Why is food so much worse in Germany than many of its neighbors? Germany certainly doesn’t have less material ability to produce good food; indeed, Germany is richer than the countries around it. There’s probably something about German food supply chains that is impoverished relative to France and Italy, but the Germans themselves don’t seem too upset about it. It just seems that the Germans are stuck in an objectively worse market equilibrium than their neighbors: the food is bad and they’ve gotten used to it. The obvious question then is where else these kinds of reflexive patterns apply, and where else we’re stuck in some objectively inferior equilibrium, even if preferences are in some superficial sense being sated.
• While this is an extremely banal and obvious point, I hadn’t until recently thought much about or internalized how much one can study reasonably objective things ("the status of women in society", say) through artwork. (Thanks to @_alice_evans for opening my eyes here.) In this vein, I’m pretty excited about the possibilities over the coming years in computational art analysis. I want something that’s conceptually similar to Google Ngram timelines but for the visual arts.
• We've always tried to do things well at Stripe. I've come to see that attempting to do them beautifully is often a helpful way to break out of standard practices and to do something with greater novelty and in a way that might have other benefits besides. (Also, excellent people want to do great work because it is intrinsically satisfying. Explicitly allowing aesthetic considerations to carry weight avoids having to justify every assessment with some kind of torturous empiricism.)
• In his Nobel Lecture, Solzhenitsyn said that, among the Platonic virtues of goodness, truth, and beauty, that beauty is special, for it possesses a unique kind of irrefutability. He notes that arguments, writing, and philosophical systems can all be predicated on misapprehensions, but that “a true work of art carries its verification within itself.” He proceeds to observe that when goodness and truth are threatened, the “ever surprising shoots of beauty will still force their way through.” There is a lot of specious and motivated reasoning in the world today and plenty of questionable value systems. I don’t think that beauty directly reflects any definitive trait, but I’m intrigued by the idea that it can be a marker of deeper metaphysical coherence.
Anton retweeted
Have you built a language model? You should. It's so much fun to chat with something you made.
Anyone can do it, too. I made an app that teaches the fundamentals and gives you everything you need to build your own: languagemodelbuilder.com
Anton retweeted
TIL that @shl runs much of his business, Gumroad, through a hermes agent called @gumclaw. A public site outlines the whole operating system they've built to have an agent run the business. Here's what stood out:
The agent runs on a dedicated mac and cron jobs wake it up to check support tickets, watch X mentions, work on engineering tasks, update finance docs, and follow up on previous work. The underlying model is Fable 5.
Each session starts new. The model itself doesn’t remember anything, so memory comes from plain-text policy files, daily logs, people notes, watchlists, sqlite ledgers, cloned repos, and an index explaining where everything lives.
The way it “learns” is cool. When Gumclaw makes a mistake, the correction becomes a dated policy rule that every future session must read. It's memory is a filesystem, and the filesystem is constantly being updated.
The agent has around 40 skills and 100 purpose-built scripts. If the task repeats, the process becomes a script or skill.
Perhaps the coolest workflow is support-to-engineering. Gumclaw can:
- pull a support ticket from helper
- verify the customer’s situation in gumroad’s internal database
- discover that the problem is actually a product bug
- reproduce it in the codebase
- write and test the fix
- open a PR
- add the customer to a follow-up ledger so another session can contact them when the fix is deployed.
The permission model is well thought out: routine work happens autonomously, broadcast actions require approval, and high-stakes decisions are escalated.
This fully thought-through and well-engineered operating system and the associated workflows are what turn the agent into something that can actually do useful work for Gumroad every day.
Worth reading!
gumclaw.github.io/how-i-work…
Anton retweeted
Bridgewater just published numbers that should make every frontier lab nervous.
The world's largest hedge fund tested Gemini, Claude, and GPT on six document filtering tasks its investors do every day. Naive prompts scored around 50%. A coin flip. Expert-written prompts pushed accuracy to 78%. Investors needed 80% before they'd trust the system in their workflow, and no frontier model cleared it. GPT 5.4 cost 43% more than 5.2 and was barely more accurate.
So they fine-tuned Qwen3-235B on Tinker instead. 84.7% accuracy. 29.8% fewer mistakes than the best frontier model. At 1/14th the inference cost.
The smartest part is buried in the middle of the paper. Their vendor-labeled training data was riddled with wrong labels, and expert labeling costs too much to run on everything. Their fix: train a model on the noisy dataset, then run it back over its own training data. Any example the model disagreed with got routed to senior investors, because either the example was genuinely hard or the label was wrong. The model's own confusion became a detector for bad labels.
Prompting hit a ceiling for a structural reason. A prompt captures only the judgment an expert can put into words. Twenty years of taste about which central bank memo actually signals a rate move doesn't compress into instructions. It transfers through labeled examples.
Every institution sitting on decades of expert decisions just learned that those archives can train a model that beats the frontier at their specific job. The alpha was in the filing cabinet the whole time.
Bridgewater used their unique financial knowledge and partnered with us on @tinkerapi to fine-tune a model that helps their analysts focus on what's important. Experts improving AI that empowers experts.
thinkingmachines.ai/news/lea…
Anton retweeted
The new business model that pays for the Internet in a world of AI agents instead of humans browsing.
Will take a while as most humans are still using AI as a siloed browser experience, but as soon as they start using AI like everyone in tech, the model will flip and quickly.
We're opening the waitlist for our Monetization Gateway, which will allow you to charge for any web page, dataset, API, or MCP tool behind Cloudflare. The charges will settle in stablecoins over the x402 open protocol. cfl.re/4eUFdt6
Anton retweeted
Expect to see a lot more of this. Vibe coding, 3d printing, cheap motors, and boards (soon, after the Great Correction) WILL lead to a true cambrian explosion of robotics
Home-built robotics projects will be the new "my first weather/calendar app" entry point for "SWE" types
An open-source robot vacuum you build yourself — Raspberry Pi, ROS 2, 2D LiDAR, Home Assistant, 3D printed chassis. No cloud, fully local.
oomwoo is early stage and building in public. The community can contribute modules in parallel — from SLAM navigation to dust bin design.
makerspet.com/blog/building-…
#ROS2 #RaspberryPi
Anton retweeted
The strongest relationships have a combination of high expectations and high support...
Anton retweeted
It's insane to think that in another ~6 months, we'll likely have Fable-level intelligence in open-sourced models.
Open-source AI is now 4 months behind frontier LLMs.
Two years ago, it was 12.
At this pace, Fable-level intelligence will be free to download by end of year.
Anton retweeted
The best way to think of Record & Replay™️ is to imagine it as a new primitive
Ask Codex to watch you do something on your computer -- anything in your day-to-day
And Codex will process the (1) recording and (2) events (same spine as computer use) to understand (A) exactly what happened, and then (B) turn it into a skill you can use for the future
Some ways I've been using R&R:
- formatting calendar invites in my own particular fashion (I had Codex watch once, and now it matches my pattern when using the gCal plugin)
- drafting PRs and sharing them to the right slack channel with the exact formatting I like
- testing onboarding flows (Codex can pickup on different friction points during the recording -- for this, I don't need it to create a skill)
Anton retweeted
OpenAI just killed the worst part of building iOS apps with AI.
The new Codex "Build iOS Apps" plugin can now:
→ run your app in an in-app browser
→ open SwiftUI previews
→ hot reload edits
all without leaving Codex.
No more copy-paste-build-screenshot loop. The agent sees what it ships.
This is the loop indie devs have been begging for.
Anton retweeted
OpenAI might swallow up every piece of software eventually.
Anton retweeted
🤯 Midjourney -- yes, the AI image company -- just shipped a brand new type of imaging machine. 🤯
- 100x faster than an MRI.
- 10x cheaper.
Full body scanned in 60 seconds instead of an hour in a tube. Ultrasound based, MRI-level resolution.
And it's real -- not a concept, a working machine. You step into a shallow pool of warm water, a ring of half a million sensors sends sound through your body from every angle, and ~60 seconds later you have a 3D map of your insides down to a fraction of a millimeter. No radiation, no tube, no lying still.
They're not even building it as a hospital machine -- they're building a spa. The scan is a side-effect of a place you'd want to hang out anyway.
Lastly, it is built by 9 people. NINE PEOPLE.
You can just do things.
