Gordon
@gordon.bsky.social
Everything around me was someone’s lifework.
Pinned
Gordon
@gordon.bsky.social
· Apr 22
self-certifying data… over http
Unironically, this makes me realize that our physical environment will be increasingly reorganized around the logic of vectors that happen to be clustered in latent space.
*OF COURSE AI-clusters "belong in outer space," not because that makes any sense technically, but because they're very closely-related stochastically in the scraped-and-harvested Internet discourse
January 10, 2026 at 7:43 AM
Unironically, this makes me realize that our physical environment will be increasingly reorganized around the logic of vectors that happen to be clustered in latent space.
Reposted by Gordon
Survival of the fittest code!
Our paper explores LLMs driving an evolutionary arms race in Core War, where assembly programs fight each other. We task LLMs with evolving "Warriors" in a virtual machine, producing chaotic, self-modifying code dynamics.
Blog: sakana.ai/drq
Paper: pub.sakana.ai/drq/
Our paper explores LLMs driving an evolutionary arms race in Core War, where assembly programs fight each other. We task LLMs with evolving "Warriors" in a virtual machine, producing chaotic, self-modifying code dynamics.
Blog: sakana.ai/drq
Paper: pub.sakana.ai/drq/
Introducing Digital Red Queen (DRQ): Adversarial Program Evolution in Core War with LLMs.
In this work, we explore how LLMs can drive open-ended adversarial evolution of programs within the Core War environment.
Blog sakana.ai/drq
Website pub.sakana.ai/drq/
ArXiv arxiv.org/abs/2601.03335
Thread:
In this work, we explore how LLMs can drive open-ended adversarial evolution of programs within the Core War environment.
Blog sakana.ai/drq
Website pub.sakana.ai/drq/
ArXiv arxiv.org/abs/2601.03335
Thread:
January 8, 2026 at 5:11 PM
Survival of the fittest code!
Our paper explores LLMs driving an evolutionary arms race in Core War, where assembly programs fight each other. We task LLMs with evolving "Warriors" in a virtual machine, producing chaotic, self-modifying code dynamics.
Blog: sakana.ai/drq
Paper: pub.sakana.ai/drq/
Our paper explores LLMs driving an evolutionary arms race in Core War, where assembly programs fight each other. We task LLMs with evolving "Warriors" in a virtual machine, producing chaotic, self-modifying code dynamics.
Blog: sakana.ai/drq
Paper: pub.sakana.ai/drq/
"Anekāntavāda ("many-sidedness") is the Jain doctrine that ultimate truth and reality is complex and has multiple aspects and viewpoints. According to Jainism, no single, specific statement can describe the nature of existence and the absolute truth." en.wikipedia.org/wiki/Anekant...
Anekantavada - Wikipedia
en.wikipedia.org
January 8, 2026 at 6:53 PM
"Anekāntavāda ("many-sidedness") is the Jain doctrine that ultimate truth and reality is complex and has multiple aspects and viewpoints. According to Jainism, no single, specific statement can describe the nature of existence and the absolute truth." en.wikipedia.org/wiki/Anekant...
THOMASINA: Do we believe nature is written in numbers?
SEPTIMUS: We do.
THOMASINA: Then why do your shapes describe only the shapes of manufacture?
SEPTIMUS: I do not know.
THOMASINA: Armed thus, God could only make a cabinet.
SEPTIMUS: We do.
THOMASINA: Then why do your shapes describe only the shapes of manufacture?
SEPTIMUS: I do not know.
THOMASINA: Armed thus, God could only make a cabinet.
January 8, 2026 at 6:40 PM
THOMASINA: Do we believe nature is written in numbers?
SEPTIMUS: We do.
THOMASINA: Then why do your shapes describe only the shapes of manufacture?
SEPTIMUS: I do not know.
THOMASINA: Armed thus, God could only make a cabinet.
SEPTIMUS: We do.
THOMASINA: Then why do your shapes describe only the shapes of manufacture?
SEPTIMUS: I do not know.
THOMASINA: Armed thus, God could only make a cabinet.
Reposted by Gordon
This reminds me of the Law of Requisite Variety -- controlling a situation (that is, reducing its variability) doesn't _eliminate_ the variability, it moves it elsewhere in the (larger) system.
January 7, 2026 at 2:56 PM
This reminds me of the Law of Requisite Variety -- controlling a situation (that is, reducing its variability) doesn't _eliminate_ the variability, it moves it elsewhere in the (larger) system.
the most ambitious crossover event in history
SFI welcomes Miller Scholar Francis Spufford.
A Fellow of the Royal Society of Literature, he teaches at Goldsmiths, University of London. His books, including Red Plenty, Golden Hill, and the forthcoming Nonesuch, explore the entanglement of historical, economic, and social forces.
A Fellow of the Royal Society of Literature, he teaches at Goldsmiths, University of London. His books, including Red Plenty, Golden Hill, and the forthcoming Nonesuch, explore the entanglement of historical, economic, and social forces.
SFI welcomes Miller Scholar Francis Spufford
Francis Spufford, an internationally award-winning writer of nonfiction and fiction, has joined SFI as Miller Scholar. A Fellow of the Royal Society of Literature, Francis Spufford teaches writing at ...
www.santafe.edu
January 7, 2026 at 6:16 PM
the most ambitious crossover event in history
Reposted by Gordon
The bots will not use physical cash.
January 6, 2026 at 4:01 PM
The bots will not use physical cash.
LLM-as-judge is unreasonably effective
Its noticing others gaining the same insight about adverserial loops improving output
www.reddit.com/r/ClaudeAI/c...
www.reddit.com/r/ClaudeAI/c...
From the ClaudeAI community on Reddit
Explore this post and more from the ClaudeAI community
www.reddit.com
January 7, 2026 at 5:47 AM
LLM-as-judge is unreasonably effective
Reposted by Gordon
this niche/targeted bluesky client that @mariaa.bsky.social and @eugenevinitsky.bsky.social are building is super interesting — it’s for researchers, by researchers…
super excited to see more new clients popping up that aren’t trying to be for everyone but rather for a specific community
super excited to see more new clients popping up that aren’t trying to be for everyone but rather for a specific community
While we also hype the moderation features, I'm really excited about the paper discovery tools @mariaa.bsky.social and I are starting to build. Open social means we can bootstrap onto existing discussions that are happening
January 6, 2026 at 1:33 AM
this niche/targeted bluesky client that @mariaa.bsky.social and @eugenevinitsky.bsky.social are building is super interesting — it’s for researchers, by researchers…
super excited to see more new clients popping up that aren’t trying to be for everyone but rather for a specific community
super excited to see more new clients popping up that aren’t trying to be for everyone but rather for a specific community
Reposted by Gordon
I am so glad Moxie is building again. We all benefit.
The architecture described here is how to use cloud for compute in a privacy-preserving way.* In this case, for a personal AI agent.
*Yes, I'm aware TEEs have been broke in the past...
confer.to/blog/2026/01...
The architecture described here is how to use cloud for compute in a privacy-preserving way.* In this case, for a personal AI agent.
*Yes, I'm aware TEEs have been broke in the past...
confer.to/blog/2026/01...
Private inference
When you use an AI service, you’re handing over your thoughts in plaintext. The operator stores them, trains on them, and–inevitably–will monetize them. You get a response; they get everything.
confer.to
January 5, 2026 at 6:47 PM
I am so glad Moxie is building again. We all benefit.
The architecture described here is how to use cloud for compute in a privacy-preserving way.* In this case, for a personal AI agent.
*Yes, I'm aware TEEs have been broke in the past...
confer.to/blog/2026/01...
The architecture described here is how to use cloud for compute in a privacy-preserving way.* In this case, for a personal AI agent.
*Yes, I'm aware TEEs have been broke in the past...
confer.to/blog/2026/01...
Bateson: "Do thermostats lie? Yes, no, yes, no, yes, no. You see, the cybernetic equivalent of logic is oscillation.”
January 5, 2026 at 1:32 PM
Bateson: "Do thermostats lie? Yes, no, yes, no, yes, no. You see, the cybernetic equivalent of logic is oscillation.”
Reposted by Gordon
Your daily reminder that in complex conflict scenarios, long-term strategy is important: arxiv.org/abs/2508.13213
AI sustains higher strategic tension than humans in chess
Strategic decision-making involves managing the tension between immediate opportunities and long-term objectives. We study this trade-off in chess by characterizing and comparing dynamics between huma...
arxiv.org
January 3, 2026 at 9:56 AM
Your daily reminder that in complex conflict scenarios, long-term strategy is important: arxiv.org/abs/2508.13213
“Rigid plans tend to frame surprises as problems”
January 4, 2026 at 5:08 AM
“Rigid plans tend to frame surprises as problems”
AI as Extended Cognition
January 3, 2026 at 5:59 AM
AI as Extended Cognition
Reposted by Gordon
This is a great example of why I say "if it can be explained in words and the output is words, you CAN get an LLM to do it."
It may not always be worth it to do it, but there isn't a moat. And consider further: what happens if you can teach this refinement process to the LLM?
It may not always be worth it to do it, but there isn't a moat. And consider further: what happens if you can teach this refinement process to the LLM?
A hypothesis is "If I change parameter X, I will see result Y." That's a specific claim that can be tested and disproven.
But if I TELL IT how to write a good hypothesis, then it DOES! Now I have a machine that can take any arbitrary paragraph and rephrase it in the form of a testable hypothesis!
But if I TELL IT how to write a good hypothesis, then it DOES! Now I have a machine that can take any arbitrary paragraph and rephrase it in the form of a testable hypothesis!
January 3, 2026 at 3:12 AM
This is a great example of why I say "if it can be explained in words and the output is words, you CAN get an LLM to do it."
It may not always be worth it to do it, but there isn't a moat. And consider further: what happens if you can teach this refinement process to the LLM?
It may not always be worth it to do it, but there isn't a moat. And consider further: what happens if you can teach this refinement process to the LLM?
Reposted by Gordon
Wikipedia is truly one of the greatest things on the Internet. It is absolutely incredible, and must be preserved at all costs.
January 1, 2026 at 10:19 PM
Wikipedia is truly one of the greatest things on the Internet. It is absolutely incredible, and must be preserved at all costs.
Reposted by Gordon
"What is the pattern that connects the crab to the lobster and the primrose to the orchid, and all of them to me, and me to you?" - Gregory Bateson
January 1, 2026 at 2:50 PM
"What is the pattern that connects the crab to the lobster and the primrose to the orchid, and all of them to me, and me to you?" - Gregory Bateson
Reposted by Gordon
What we communicate are models of the world, not the world.
October 31, 2025 at 4:04 PM
What we communicate are models of the world, not the world.
In the quietness, with nothing to accomplish, there is only direct experience.
(Dogen Zenji)
(Dogen Zenji)
If you are too confident in your representational maps of reality you will miss seeing reality
December 30, 2025 at 12:58 PM
In the quietness, with nothing to accomplish, there is only direct experience.
(Dogen Zenji)
(Dogen Zenji)
“How rare to be born a human being!”
I made a birth lottery roller, inspired by Giving What We Can's, but inclusive of nonhuman animals!
birth-lottery.thetetra.space
birth-lottery.thetetra.space
The Birth Lottery
Draw again from the birth lottery to take a different sample from the way the circumstances outside your control go for living beings on Earth today.
birth-lottery.thetetra.space
December 29, 2025 at 7:43 AM
“How rare to be born a human being!”
The the treatment for severe acute malnutrition was invented by accident after a scientist got exasperated at his kids for eating too much Nutella en.wikipedia.org/wiki/Plumpy%...
December 28, 2025 at 6:35 PM
The the treatment for severe acute malnutrition was invented by accident after a scientist got exasperated at his kids for eating too much Nutella en.wikipedia.org/wiki/Plumpy%...
(I need to get this piece published)
December 28, 2025 at 5:04 PM
(I need to get this piece published)
Reposted by Gordon
"I collected seventeen ways in which unsought findings have been made."
...
Child, student or outsider
Playing
Joke
Dream or "forgetting-hypothesis"
...
"
...
Child, student or outsider
Playing
Joke
Dream or "forgetting-hypothesis"
...
"
“Taxonomy of luck”
Windfalls or wind thefts
Unforeseeable lost or gained opportunities
Accidents
Narrow escapes
Coincidences
Consequence-laden mistakes in identification or classification
Fortuitous encounters
…
www.stat.berkeley.edu/~aldous/Real...
Windfalls or wind thefts
Unforeseeable lost or gained opportunities
Accidents
Narrow escapes
Coincidences
Consequence-laden mistakes in identification or classification
Fortuitous encounters
…
www.stat.berkeley.edu/~aldous/Real...
Taxonomies of chance, risk and unpredictability.
www.stat.berkeley.edu
December 28, 2025 at 8:41 AM
"I collected seventeen ways in which unsought findings have been made."
...
Child, student or outsider
Playing
Joke
Dream or "forgetting-hypothesis"
...
"
...
Child, student or outsider
Playing
Joke
Dream or "forgetting-hypothesis"
...
"
“Taxonomy of luck”
Windfalls or wind thefts
Unforeseeable lost or gained opportunities
Accidents
Narrow escapes
Coincidences
Consequence-laden mistakes in identification or classification
Fortuitous encounters
…
www.stat.berkeley.edu/~aldous/Real...
Windfalls or wind thefts
Unforeseeable lost or gained opportunities
Accidents
Narrow escapes
Coincidences
Consequence-laden mistakes in identification or classification
Fortuitous encounters
…
www.stat.berkeley.edu/~aldous/Real...
Taxonomies of chance, risk and unpredictability.
www.stat.berkeley.edu
December 28, 2025 at 8:29 AM
“Taxonomy of luck”
Windfalls or wind thefts
Unforeseeable lost or gained opportunities
Accidents
Narrow escapes
Coincidences
Consequence-laden mistakes in identification or classification
Fortuitous encounters
…
www.stat.berkeley.edu/~aldous/Real...
Windfalls or wind thefts
Unforeseeable lost or gained opportunities
Accidents
Narrow escapes
Coincidences
Consequence-laden mistakes in identification or classification
Fortuitous encounters
…
www.stat.berkeley.edu/~aldous/Real...