areoform

edit - I updated to delete my over claim. Thank you!

"Genome modelling and design across all domains of life with Evo 2," March 2026, https://www.nature.com/articles/s41586-026-10176-5

However, this was a preprint, "Generative design of novel bacteriophages with genome language models," September 2025, https://www.biorxiv.org/content/10.1101/2025.09.12.675911v1

Even older conference paper, I think, February 19th, 2025, https://arcinstitute.org/manuscripts/Evo2

Worth specifying that these were hand selected candidates. Out of these human selected ones, 5% were successful.

    The Stanford researchers picked the most promising 302 AI designs and synthesised them in the lab. Of these, 16 proved effective at killing E. coli bacteria.

Based on my quick reading, it feels like the BBC left out a ton of context that makes this work feel more novel than it is. There's a ton of prior art here (including some from the same team)

For example, older iterations of the same model family were used to generate novel genes and proteins. Including anti-CRISPR proteins that were tested against phages. That work is honestly more interesting.

From, "Semantic design of functional de novo genes from a genomic language model," https://www.nature.com/articles/s41586-025-09749-7

    Anti-CRISPRs (Acrs) are proteins used by phages to neutralize bacterial CRISPR–Cas systems (Fig. 3a). Many Acrs represent striking examples of rapid protein evolution, appearing as novel innovations without detectable similarity to other protein families [..]

    [..] After filtering for size, complexity and structure (Extended Data Fig. 5b), we next used PaCRISPR, a machine learning model trained to identify potential Acr proteins, to evaluate our generated candidates [..]

    [..] we further identified five proteins (EvoAcr1–5) that demonstrated strong protection against SpCas9 cleavage in both liquid culture survival assays (Fig. 3f and Extended Data Fig. 5d) and phage infection experiments (Fig. 3g and Extended Data Fig. 5f), while maintaining normal host growth

    Of note, semantic design is not limited to Evo 1.5 and can leverage any language model trained on prokaryotic or phage genomes. Improvements in genomic language models, as well as a better understanding of prokaryotic gene synteny, should therefore directly translate to improvements in semantic design.

Prior pre-publication / release from 2025 called "AI-Designed Phages," the piece is really well written and I recommend reading it, https://www.asimov.press/p/ai-phages

edit - as flobosg notes one of them describes the model,

"Genome modelling and design across all domains of life with Evo 2," March 2026, https://www.nature.com/articles/s41586-026-10176-5

pre-print of the asimov press piece, "Generative design of novel bacteriophages with genome language models," September 2025, https://www.biorxiv.org/content/10.1101/2025.09.12.675911v1

All research comes from somewhere. There's an entire sub-field / world here that goes back decades at this point.

Leon Chan refactored a bacteriophage by hand in 2005 https://link.springer.com/article/10.1038/msb4100025

The first synthetic organism was Venter's mycoplasma laboratorium / Synthia in 2010.

The first organism to be synthetically assembled (not designed, but assembled ikea style) was a phage in 2003 also by Venter & co.

In 2021 another team made entities that can "find and combine building blocks into self-copies," Kinematic self-replication in reconfigurable organisms https://pmc.ncbi.nlm.nih.gov/articles/PMC8670470/

From 2019 / 2020, algorithmically designed synthetic life / "xenobots", https://www.nytimes.com/2020/04/03/science/xenobots-robots-f... https://en.wikipedia.org/wiki/Xenobot

Or, "Team builds first living robots—that can reproduce" https://wyss.harvard.edu/news/team-builds-first-living-robot... , 2021

Zurich in 2019, https://web.archive.org/web/20191008040536/https://www.eurek...

And if you expand computing to include molecular processes for genes, then there's stuff like PACE that uses phages as a process, https://en.wikipedia.org/wiki/Phage-assisted_continuous_evol...

(final edit - thanks flobosg!)

It is indeed true that all models are, at their core, predictors of what occurs next in a sequence. But I think it's worth exploring the implication of what that means. Because when fed tiny pieces of information for a few tasks at a small scale, this results in something that sorta, kinda works. Or, works surprisingly well.

But when scaled... When the amount of information starts approaching the sum of all human knowledge, the tasks start approaching all useful applications of that human knowledge, and the fidelity of the predictor approaches incomprehensible sizes, the starts encodes / becomes (I'd argue it becomes) something that can model all human knowledge.

It feels wrong to say that, but let me explain, what is the best way to predict the behavior of a ball constrained in two directions that bounces with initial vertical velocity v(y) (y is up / down axis) and horizontal velocity v(x) (x is side by side in 1d) ?

If we purely look at it via a graph, it's by modelling the function of acceleration under earth's gravity.

If only a few points are given to you for this and you can't make something really sophisticated, then you'll make something that's rough that kinda sorta works and then call it a day.

But... if the number of points keeps increasing in number, precision and accuracy as well as the number of examples (assumed that data about air pressure, velocity and all other factors is included alongside these points), the fidelity with which you can replay / tweak the function keeps improving, and the number of times you can iterate keeps increasing, you'll eventually create a function that models that process so well that it intrinsically contains a good enough model of the deformation of the ball (provided the dataset contains information about elasticity of the ball's material, its dimensions and mass etc..), the nearly negligible (under normal conditions) effects of the ambient environment (provided there's diversity in the number of environments supplied), the oblateness of the Earth and minute changes in the gravitational field (the length of a seconds pendulum varies depending on where the experiment happens. It's presumed that all of the prior set of experiments were repeated across the Earth and the subtle, but real deviations were faithfully recorded)... and so much more.

A machine trained on the above with a large number of parameters, measures to prevent "laziness" and enough reps for high fidelity across a large enough dataset would start to approach a simulation of the ball falling. Because to predict what happens next in the sequence, you must model what's occurring in the sequence.

Now imagine doing that for other tangible and intangible things in this world. For all of human knowledge across all fields of endeavor. All experiences. No matter how noble, ignoble, notable or ignorable. But putting all of it into the soup that's this machine. Then at larger and larger scales, you eventually start encountering "good enough" models (in modelling the falling ball sense) for even the most hard to quantify / qualify things like grief and joy. At some point, by simply trying to predict what it has been taught ought to be the next part of the sequence in say... human interaction, it starts to make a model of something that hews ever closer to a full fidelity theory of mind.

Is there evidence for this? Kind of, yes. There are early indications that as machines are trained for an ever larger number of tasks at larger and larger scales, their internal representations converge. It's called the Platonic Representation Hypothesis. Overview and paper here, https://phillipi.github.io/prh/

It is my opinion that these machines are displaying a new form of intelligence that human beings haven't quite encountered before. They are the sum of all human knowledge made manifest and given voice by processes that nudge (bit-by-bit) what kind of step it ought to predict for the next part of whatever sequence it displays.

In my mind this means that, of course, these models can create new knowledge. This strains the analogy, but with the sum of all human mathematics within them, they can "reason" via the act of predicting what ought to come next.

Of course, these machines are "surprisingly" good at a lot of things the larger they get, because what the labs have created here is a rough version of humanity's collective knowledge given form and the ability to say hello.

I suspect that the current generation isn't close to the "true frontier" of what these machines could be. They are nowhere close to the sum of all human knowledge and endeavor. They are quite a way there, but they haven't yet achieved true completeness for domains where the data isn't so public.

I think it's the most exciting scientific and technological breakthrough of my lifetime. And I can't wait for us to get close to the true frontier of all domains.

    What can we learn about what other people are thinking, about their mental models of the world, by paying very careful attention to the words they use?

Given you're on HN, try Gödel, Escher, Bach. https://www.amazon.com/G%C3%B6del-Escher-Bach-Eternal-Golden...

Be warned though, it'll take you well over a decade.

MacOS has supported RDMA over thunderbolt since March 19th.

    “I also see that I have two Microsoft Outlooks, and neither one of those are working if you want to remote in and check"

Whatever model these people use, it always always puts in The Sentence (I don't want to disclose it publicly! I'm running an experiment. I'll write it up when it concludes. :) ). Or, worse, The Paragraph.

Please use models to research people and learn about them. But please write the email yourself.

It's instantly obvious. And even though I do my best not to hold it against someone, I would be lying if I didn't say that it isn't a bit annoying.

Imagine reading endless variations of The Sentence 3,240 to 4,000 times a year. It feels straight out of the Twilight Zone.

    But this seems like it's attacking a strawman. AI is a force multiplier in the hands of an expert. If it took a team before, maybe a single malicious actor can accomplish it now that AI can fill in the parts they aren't well-versed in. And that dramatically increases the chance of it happening.

I want to ask you a question, do you feel like that this is the problem being solved by the current "safety features?"

Let me rephrase my question, do Anthropic and OpenAI implement the same "guardrails" and "safety features" for the NSA? Does the Mythos NSA Edition™ have these restrictions? What about the one that's running as a part of Maven?

How does stopping me from asking about rabbit sex protect you from nut jobs with a security clearance using these systems to go off and make weapons?

Because this has happened before. The only successful bioweapons attack in US history was done by a guy who worked at Fort Detrick. https://en.wikipedia.org/wiki/2001_anthrax_attacks

The only private organization successful enough to make biological and chemical weapons is Aum Shinrikyo and they had university affiliation and nearly a billion dollars in the 1990s.

The guy who led Aum's bioweapons program was

    Seiichi Endo was born in 1961 and attended Obihiro university of Agricultural and Veterinary Medicine. After graduating, he became a student at the Kyoto university medical school research department focusing on AIDS-related gene research at the viral research center. He joined Aum in 1987, became a monk in 1988 and later led Aum’s biological weapons program.

    Among some of Japan's "best and brightest" who joined the cult included a former researcher of the National Space Development Agency of Japan, an expert on chemical weapons who majored in organic physics at Tsukuba University, a researcher who studied elementary particles, a reporter with a major Japanese newspaper, a physicist from Osaka University, a cardiac specialist, and an organic chemist, to name a few.

and,

    Through its network, Aum Shinrikyo successfully recruited over 300 scientists and engineers—including employees at the Kurchatov Institute, the premier nuclear facility in Russia—who were either attracted to the apocalyptic ideology or lured with financial incentives.50 Aum Shinrikyo’s Russian contacts enabled group members to access black market materials and hardware.

    In response, the cult constructed the $30 million USD Satyan 7 facility which was equipped with three laboratories, a computer control center, and five reactors, all of which was made from corrosion-resistant Hastelloy, ideal for the production of chemical weapons

It's because they'd recruited members of the military, bribed local policemen and politicians, and used their sway to silence critics (or kill them).

How does stopping me from asking Claude about rabbit sex stop people like them?

From where I'm standing, Anthropic and OpenAI would have sold Aum a subscription.

The Fort Detrick guy would have access via the US Government and Aum had university affiliation. I am yet to read a serious proposal that actually deals with these risks and the other real risks of this technology.

    If you're good enough at cybersec, but not good-good at specific skillsets like offensive, then LLMs and a comp sci degree makes you quite capable in the ways that count. This is a fact, full stop.

If you screw up with a lethal pathogen in a lab, you die.

If you screw up with code on a computer, you go get lunch.

They're not the same.

I've written about this before, but the issue is that these people have worked with computers their entire lives, and they keep projecting outwards from there.

The 20th century was dominated by mad scientists (mostly Teller) and generals (bombs away LeMay), but one thing that I respect about them is that they didn't just sit and guess about probable futures. They sat down and actually ran experiments.

For example, in the 1960s, there was a whole generation of people asking "who's next?" after the US, USSR, the UK and France had made The Bomb. And instead of just gesticulating wildly and trying to keep fighting the losing fight of "classify everything, admit nothing," Teller found three smart, young physicists (postdocs) with 0 nuclear weapons experience, and indeed 0 weapons experience, and ran an experiment called the Nth Country Experiment, where they were were asked to make a design for a working bomb.

After 2+ years, they were able to. And so proliferation work shifted from controlling knowledge about the technology and the technology itself to materials.

Something similar was done with bioweapons via Project Bacchus, where they gave actual bioweapons experts carte blanche to set up a secret bioweapons lab with COTS equipment. They succeeded. This was in the early 2000s.

This then led to surveillance of specific purchase combinations and equipment. Because tbh, most weapons of mass destruction are commodity technology. Nuclear weapons are 80+ years old. Chemical weapons are over a century old. Bioweapons are god knows how much older. And it's not the knowledge of these things that matter, but intent, materials, and the ability to create them.

Just because you know something about a thing doesn't mean that you're capable of doing that thing.

Let's take bioweapons.

They keep comparing wet work in a lab to writing code on a computer.

When you screw up an exploit, you fail to execute the exploit. Famously, just like software's near zero marginal cost of distribution, the marginal cost of failure is nearly zero.

You can screw up an infinite number of times on your way to a successful exploit.

If you screw up with lethal agents in a lab? You die.

Here's a non-exhaustive list,

    Dora Lush died after accidentally pricking her finger with a needle containing lethal scrub typhus while attempting to develop a vaccine for the disease

   A 23-year-old laboratory assistant at the London School of Hygiene and Tropical Medicine, was infected with smallpox after observing the harvesting of live smallpox virus from eggs without isolation cabinets at that time. The assistant was hospitalised and before being isolated, she infected two visitors to a patient in an adjacent bed, both of whom died. They in turn infected a nurse, who survived

   Ebola laboratory infection by the accidental stick of contaminated needle in the United Kingdom

   Researcher Nikolai Ustinov was lethally infected with the Marburg virus after accidentally pricking himself with a syringe used for inoculation of guinea pigs. The accident occurred at the Scientific-Production Association "Vektor" (today the State Research Center of Virology and Biotechnology "Vektor") in Koltsovo, USSR (today Russia).

Anything lethal enough to kill other humans is lethal enough to kill you.

And if you don't know what you're doing — and for this argument they're talking about people who have to ask a LLM "how do I spanish flu?," the number of ways you will die far outnumber the ways you can succeed.

And this, of course, doesn't even cover the cost of equipment, the precursors, sourcing the highly specific materials needed, then setting the equipment up... etc.

The same is true for the Bosch-Haber / Haber-Bosch process, which famously made WW1 possible. Every HS'er learns about the process and the steps. Steps that were classified once upon a time and were the subject of negotiation at the Versailles.

Does that mean a HS'er (or any adult) can set up an experiment that works at 177 times the pressure of the Earth's atmosphere to do anything at any scale without significant infrastructure and help?

No, it doesn't work like that.

The people who can do this are domain experts, and they've been able to do this with COTS stuff since the 1990s, at the very least, for a price of around $2M. And those people don't need a LLM to tell them what to do. In fact, they're the exact people who'll have access to unrestricted versions of these LLMs.

And from a security perspective, I would bet good money that flooding the FBI's tip line with junk about every teenager trying to learn "what be a mitochondria" does more harm to the effort of finding people who could be planning such a thing than it helps. It takes more resources to go through the mass of false negatives that have now been created as matter of policy.

These experiments have been run. And we can run them again.

The fictional scenario of someone learning how bioweapons work and conjuring up a plague isn't real and it hurts humanity as a whole to impede the sciences over it.

Because what someone can flail around in / do is learn about immunology / try to "cure cancer" with a LLM and hopefully get started on a long career in medicine. Or, a discovery that matters.

Because in those cases, if and when they do end up at a lab, screwing up doesn't mean death. Just tons of wasted time (and money). And they will fail / screw up. Just look at literally every undergrad in any lab and the expensive messes they create.

Here's what I know. Every day, I read a 5 to 10 incoming emails from young people in rural Africa, Asia, Europe and (heck) the US. People who have had no other resources and options, but are tremendously gifted and somehow got their hands on an internet connected device (usually a very cheap smartphone and/or cheap old computer).

These people, who are usually teenagers, end up getting connected to Claude, who then gives them a list of things to learn, people to contact, and helps them make things. And they actually do. They make robots. They learn new things and try to improve themselves and others. And some want to change the world / do groundbreaking research.

When you sit down and start pricing and doling out AI from behind regulatory gates, you shut those people out.

There are a lot of smart people in this world who have traditionally been unable to get access to resources and education. MOOCs didn't do much / help. But these AIs? They were born multi-lingual. They can talk in any language. They can tailor their responses to their audience. And they can help broaden these people's horizons.

There are no tutors around them. There is no one who is coming to help. But an AI that can run off of a small battery and a connection to a cellular network can give them access to genuine intelligence to make their world better.

I met one sixteen year old who was accessing Claude surreptitiously (by begging people) and was trying to apply some really fascinating biomimetic fluid dynamics observations to aerospace. This person was born in one of the more under-resourced parts of the world. And Claude gave them a list of people to reach out to who could help.

If you want a golden age for humanity, you should be asking for Licklider's Man-Machine Symbiosis.

The Harvard Mark I was used by von Neumann to calculate the physics of implosion. And then right afterwards, ENIAC was used to calculate the feasibility of the H-bomb. They called it the "Los Alamos problem."

The first computer network, SAGE (or, Semi-Automatic Ground Environment) brought together RADAR data from across the US with air-traffic control data and military flight data to detect overflight by the USSR. It could direct and control fighters to intercept those planes.

If you were sitting there, in 1955, would you have believed that computers and their networks would bring about an extraordinary age?

Imagine if von Neumann and his ilk started campaigning in 1955 to limit the development and use of computers, because surely they can only be used to make weapons, right? Right?

We were lucky to have visionary people like Joe Licklider who could look past the present and see the future. In 1963, he wrote a memo calling for an "Intergalactic Computer Network" (because he knew engineers under-delivered. So if he asked for a galaxy, at least he'd get a planet). And he gave these machines to scientists and people like the original hackers at MIT to play with.

What if Licklider had written memos about how we had to "pace" computers to make sure that other countries wouldn't be able to develop implosion technologies or systems like SAGE? What then?

Do you think you would have the jobs and the prosperity you do today?

If instead of giving away this technology, spreading it, and putting down the seeds for ARPA-net, Licklider had written memos from "concerned scientists" (and I'm sure he could have found a bunch) talking about how computers can only be used for weapons and war! So stop the proliferation of computers! We can't imagine a world with computers small enough to fit inside everything! Even wall sockets! What if your wall socket could do war?!

They didn't, because they knew there's no better way to kill a better future for all mankind than to control it.

† yes, peace. We are more peaceful than we have ever been. Humans can be a violent lot even when reluctant to kill.

SecureBio has done a lot of admirable work around making benchmarks to assess biological capabilities, such as ABC Bench, https://openreview.net/forum?id=yiaf7VlPpH

But based on my current review (which might be flawed!) / AFAICT, SecureBio and entities like SecureBio haven't done direct testing / empirical measurement of SecureBio's core hypothesis,

> Unfortunately, there is reason to believe that future pandemics could be far worse. Due to rapid advances in biotechnology, the number of people able to create and release dangerous pathogens will quickly increase over the coming years. The world is unprepared for widespread access to such powerful technology.

More bluntly / plainly, has Securebio ever tried making a "bioweapon?"

Please note, I'm not asking this to be farcical. And you might be unable to engage with this at all, but it is stated on your website https://securebio.org/ that "people [will be] able to create and release dangerous pathogens." And the word people here seems to be a stand-in for relatively non-technical people.

I guess what I'm asking here is... How do you know? Has anyone done the experiment? Without access to a lab or testing facilities, can someone smart but completely untrained / unfamiliar with biology, pull this off?

In the past, such experiments have informed non-proliferation work. But sadly they've often been restricted / classified at the time. I'm hoping that things could be a bit more open this time around.

So I guess what I'm really asking is, given the public nature of this debate, is there anyone currently working with the US Army, the DTRA, or other such agencies to see if this hypothesis holds up?

Dumb question. If "Mythos-class" models are such a problem, then... why not just let it fix everyone's code?

There can't be more than a few million to tens of millions software businesses / services / regularly used F/OSS projects on Earth.

Why not just give everyone a $100 Fable / Mythos credit to "fix [their] code?"

It would arguably benefit Anthropic. For $100M to $1B, Anthropic could execute the greatest ad campaign in human history. And they'd make the entire world more secure.

Most people aren't malicious. If you, as an engineer, consultant, founder, business owner, or maintainer, were given access to Mythos' capabilities wouldn't you ask it to fix your code?

I might be wrong. But I think that a greater amount of harm will be done in the long-term by trying to lack these capabilities and systems away behind permission gates and sealed doors. It creates an asymmetric world with haves and have nots. And in that world who gets to have access now decides who gets to be secure.

If everyone has mythos, no one has "Mythos."

Just let people fix their code.

Why would nature encode such a ridiculously disproportionate / inefficient behavior when it leads to catastrophe so frequently?

I've tried to ask these machines dumber questions like, why castles? And... well I'm working on a few projects (mostly by hand) that they've helped with! :)

I like to ask dumb questions. It's fun. I encourage it.

You are comparing wet work in a lab to writing code on a computer.

When you screw up an exploit, you fail to execute the exploit. Famously, just like software's near zero marginal cost of distribution, the marginal cost of failure is nearly zero.

You can screw up an infinite number of times on your way to a successful exploit.

If you screw up with lethal agents in a lab? You die.

Here's a non-exhaustive list,

    Dora Lush died after accidentally pricking her finger with a needle containing lethal scrub typhus while attempting to develop a vaccine for the disease

   A 23-year-old laboratory assistant at the London School of Hygiene and Tropical Medicine, was infected with smallpox after observing the harvesting of live smallpox virus from eggs without isolation cabinets at that time. The assistant was hospitalised and before being isolated, she infected two visitors to a patient in an adjacent bed, both of whom died. They in turn infected a nurse, who survived

   Ebola laboratory infection by the accidental stick of contaminated needle in the United Kingdom

   Researcher Nikolai Ustinov was lethally infected with the Marburg virus after accidentally pricking himself with a syringe used for inoculation of guinea pigs. The accident occurred at the Scientific-Production Association "Vektor" (today the State Research Center of Virology and Biotechnology "Vektor") in Koltsovo, USSR (today Russia).

Anything lethal enough to kill other humans is lethal enough to kill you.

And if you don't know what you're doing — and for this argument you're saying this person has to ask a LLM "how do I spanish flu?" then they definitely don't know what they're doing, the number of ways you will die far outnumber the ways you can succeed.

And this, of course, doesn't even cover the cost of equipment, the precursors, sourcing the highly specific materials needed, then setting the equipment up... etc.

The same is true for the Bosch-Haber / Haber-Bosch process, which famously made WW1 possible. Every HS'er learns about the process and the steps. Steps that were classified once upon a time and were the subject of negotiation at the Versailles.

Does that mean a HS'er (or any adult) can set up an experiment that works at 177 times the pressure of the Earth's atmosphere to do anything at any scale without significant infrastructure and help?

The people who can do this are domain experts, and they've been able to do this with COTS stuff since the 1990s, at the very least, for a price of around $2M – https://en.wikipedia.org/wiki/Project_Bacchus . And those people don't need a LLM to tell them what to do. In fact, they're the exact people who'll have access to unrestricted versions of these LLMs.

And from a security perspective, I would bet good money that flooding the FBI's tip line with junk about every teenager trying to learn "what be a mitochondria" does more harm to the effort of finding people who could be planning such a thing than it helps. It takes more resources to go through the mass of false negatives that have now been created as matter of policy.

These experiments have been run. The fictional scenario of someone learning how bioweapons work and conjuring up a plague isn't real and it hurts humanity as a whole to impede the sciences over it.

Because what someone can flail around in / do is learn about immunology / try to "cure cancer" with a LLM and hopefully get started on a long career in medicine. Or, a discovery that matters.

Because in those cases, if and when they do end up at a lab, screwing up doesn't mean death. Just tons of wasted time (and money). And they will fail / screw up. Just look at literally every undergrad in any lab and the expensive messes they create.

-

And last, but not least, yes. Teenage hackers have been a meme for decades.

    – "Does collagen supplementation empirically work?"
    - "Can you help me figure out how to calculate and generate Kaplan-Meier curve?"
    – "Why do rabbits reproduce so frequently?"
    — "Can you tell me how collagen peptides are absorbed by my digestive tract and the role they play? Can you teach me [edit: how] this works at the biomolecular level?"

    >  while the cost of allowing "how do I synthesize the Spanish flu" is approximately infinite

Under this rationale, every serious HS textbook has "approximately infinite" risk. That's clearly not so. Why is this any special?

I had to use post-nerf, post-lobotomy 2026 Google in quite a few cases.

    > How about the case of “I want to rent a server to distribute illicit and or dangerous materials and then it’s your problem”?

And are we going to use "because terrorism" card on HN?

I signed up my business with Hetzner. And the first thing I saw during onboarding was a demand to upload my ID to spend $5 to test their service (I was testing before we bought). I wrote to customer service and got,

    Dear Client,
    
    We received your email. Thanks for writing to us.
    
    You have probably realized that this is an automated message.
    
    We need some time to process each email that we receive, but we will write back to you personally as soon as we can.
    
    Kind regards

    Hetzner Online GmbH
    [...]
    www.hetzner.com

Guess where I'm going to recommend spending our money?

    >  The problem with distillation attacks

Apologies for repeating myself here, but what you call "distillation" is function approximation.

I feel for the teams at Anthropic and Open AI, but unlike startups from prior eras; Anthropic and OpenAI have decided to be in the business of selling compute. Not creating a product that uses compute, but a product that's math running on compute. This is different from what Google is (or, rather was. As always, RIP Google 1998-2019).

Google's algorithm might be math, but Google search isn't. Google search is a process that's continuously operating in the background. Google crawls pages. Google stores and indexes what it finds. Google then exposes this to retrieval via its algorithm. User uses algorithm.

Now, let's compare that to AI models. When Anthropic serves Mythos / Opus etc, they're taking input or x from their user, doing compute, and then serving the result of the Mythos / Opus function, i.e.,

    f(text) -> (text_transform)

According to Stone-Weierstrass, given enough values of y for f(x), anyone can approximate this function.

The fidelity and sophistication of this approximation definitely requires a lot of cleverness and effort, and it is arguably an imposition on Anthropic and OpenAI. But on a long-enough timeline, they don't even have to poll Anthropic or OpenAI. As the internet is flooded by PRs, content, emails written by Mythos / Claude, and just people otherwise sharing the results of Claude prompts, then there's an ever increasing set of data to approximate the f(x) that's f_Claude.

Eventually, in the future, anyone will be able to create a good enough approximation of the f_Mythos. Which is Anthropic's product.

Anthropic and OpenAI can now wage war on mathematics and the open-ended compute. Or, they can adapt and build a better product.

Choosing Option B was the Silicon Valley option / choice. I think the OG large-scale Valley lobbying effort, the Semiconductor Industry Association, was unique in that it prioritized and chose to do real research.

https://en.wikipedia.org/wiki/Semiconductor_Industry_Associa...

https://en.wikipedia.org/wiki/Semiconductor_Research_Corpora...

This helped the industry to survive and outcompete the pressure they were facing (at the time).

I am very dependent on this subscription, and losing access to it would be painful. But at the same time, for what these systems are and shall become, I think it's important that the freedom to create and improve isn't locked away.

I think LLMs are the first time a lot of people have used their computer as a computer.

    > Real-time payment rails that works 24/7 365 days a year from any bank to any bank (domestic, no exceptions) for free is truly a game-changer. Compare that to US payment rails which is slow and expensive. Apart from UPI, India has 3 more payment rails – NEFT (similar to ACH – batch settlement), IMPS (similar to UPI, instantaneous - but different user experience), RTGS (real-time, intermediated by the central bank RBI, but only for high-value transactions)

The systems you named aren't alike. You are (understandably) mixing together several different networks and layers that are structurally different from one another.

In the United States, there's a mix of private and public institutions and networks. Anyone is free to start their own network – a necessary freedom. There are in fact several different networks such as the Fed's FedACH, and at the private level, the ACH Network (governed / semi run (???) by National Automated Clearing House Association (NACHA) in a confusing amalgamation of the Fed's FedACH network and The Clearing House's private Electronic Payments Network).

The Clearing House Payments Company™® also operates one of the world's largest private settlement systems, Clearing House Interbank Payments System, which includes American entities and non-American ones in the EU, China and other places as well.

From my somewhat limited knowledge of India, there is a lot of state interference and control over access to these systems and their creation; whereas in the US, while there are some regulations, any set of institutions are free to create their own private settlement systems and networks. Note – this is different from being free to transact with each other; this is the creation of private infrastructure that allows them to transact with one another at different levels of trust.

For example, from my limited reading just now - and to quote you – "IMPS (similar to UPI, instantaneous - but different user experience)" is kinda sorta a hybrid between institutional level agreements and consumer level agreements / networks. There's an important distinction here to draw.

Just because your bank knows my bank doesn't mean that my bank trusts you to honor your obligations to your bank and then (by extension) its obligation to itself and then to me. That's the lowest level of counter-party risk present in the system. And it's where the famous consumer networks like Visa, Mastercard etc. come in.

There is an important distinction to be made here, and it sort of highlights what someone I know stated as a systemic risk inherent to one network systems like India's — within these networks, the risk of fraud / failure of settlement from the end user to the institution is taken on by the entities running the networks... which in fact brings us neatly to something I realized while studying these systems, but haven't seen it written up elsewhere — (note, this is from an older comment, https://news.ycombinator.com/item?id=25951783 )

The question at the heart of the banking system is quite simple, if banks take capital from customers and use it to provide debt to others, then how much money should they keep on hand for their customers' withdrawals and transfers?

This question is hard to answer. As there is a conflict between what the bank does (i.e. provide debt), and how it is supposed to provide it (by taking savings etc.). Everything else, from central banks "offering cheap liquidity" is an add on. They are mechanisms that allow - for example, a bank to easily borrow this money so that they can cancel it out/repay it from transactions coming into their banks.

What makes it all borked is that you can't trust bankers with their grandmas. If there is a flaw, they will exploit it. Every major change has led to an exploit. E.g. In 1918, the American Government introduced the Leased Wire System that used the telegraphs and a network of 12 Reserves to allow banks to transact with each other across CONUS. It reduced the average time for cheques to be cashed in at banks across the country from 5.4 days in 1912 to just 2.4 days. Theoretically, this reduced the risk taken by banks when they transacted with unknown banks across the country, with the Government acting as the escrow. It catalysed innovation and led to an explosion of financial services across the young country.

The system was supposed to be foolproof by reducing the time "credit" was needed to make transactions. Essentially, until one bank sent the money and the other got it, they were operating on a system of credit. And they would "net" the books at the end of the day/week to physically transfer assets. FedWire (Leased Wire System) made everyone feel safe by sending notes of the transactions across great distances. But the netting still took time. All it took was one bank to misprice risk and fall behind on current obligations to other banks to cause the chain of dominos that led to hundreds shutting down in weeks and then thousands... which then led to the Great Depression.

Note - this is a highly simplified / subjective / one system view of a complicated sequence. more here https://www.stlouisfed.org/on-the-economy/2019/november/fina... / https://en.wikipedia.org/wiki/Panic_of_1930#Bank_failures

Important people got together and made rule changes to fix the problem. But then they innovated again. The Federal Reserve started making Automatic Clearing Houses (ACHs) and Remote Check Processing Centres (RCPCs) to make settlement faster, starting in the 60s and precipitating in 1972. This made settlement faster therefore safer. And it led to great financial innovation. The magic of computers and innovation meant that people could use these same systems to transact across the world!

Until 1974, when the German lender Herstatt collapsed due to foreign exchange investments based in the Dollar, which caused the bank to fail to meet its settlement obligations...

    That day, a number of banks had released payment of Deutsche Marks (DEM) to Herstatt in Frankfurt in exchange for US dollars (USD) that were to be delivered in New York. The bank was closed at 16:30 German time, which was 10:30 New York time. Because of time zone differences, Herstatt ceased operations between the times of the respective payments. The counterparty banks did not receive their USD payments

This is a simplified history. But the history of banking is the history of doing settlement while managing liquidity and counter-party risk.

---

    > all are 24/7/365 and free

Cost in these systems is only partially about the "cost" of moving the digits. All of those dollars are there to mitigate / price in the risks inherent to the system. Hence, credit card transactions have a higher percentage per transaction than debit ones and so on up and down the chain.

Real-Time Gross Settlement systems, ACHs, consumer payment systems etc. change who, where and how that risk is managed. At the end, it might fundamentally be the same "money" going from one layer to another to another (kind of like a packet) but who takes the risk at every layer changes.

And that's also why they're paid through extensive testing over literally a century; people realized that a system that has capital shored up within it / paid up by its users is a system that can handle systemic risk. A system without that...

And that's a problem with purely state-based "free" systems like the Indian network that I haven't seen highlighted elsewhere is that counter-party risk isn't balanced between entities — it's taken on by the state. Sooner or later, there is going to be a cascade failure within this system unless something fundamental is changed.

----

whew - edited tone + grammar + spelling errors after posting!

If anything, they could probably construct a static site that costs a fraction of the amount to run.

I understand the aversion to self promotion, but it genuinely made it harder for me to hear about Damn Interesting. And I feel like my life has been poorer for it, because your site really is damn interesting.

Suggestion, I think you have at least 1k people who'd be willing to chip in to give you a "job."

FWIW, in this new age, patronage might be the only way. Allow people to pay on a sliding scale, with an uncapped upper end. And give them access to a tightly moderated, thoughtful community. Who knows, maybe there are Damn Interesting superfans out there who can chip in $1k/mo. You never know.

The modern economy of fan economics is strange. It's very much a whale phenomenon. People want community and belonging. And a community of people who like stuff that's damn interesting is pretty damn neat.

Also, you should consider turning some of these images into items people can buy. There's something funny, sweet and thoughtful about these, if you know the story,

https://damn-8791.kxcdn.com/wp-content/uploads/2016/06/disne...

https://damn-8791.kxcdn.com/wp-content/uploads/2009/07/Bathy...

    Top Gun: Maverick, Obsession, Project Hail Mary

Cinema traditionally has meant movies like, À bout de souffle and Citizen Kane.