Cantrip README
the extensible, customizable, self-documenting, real-time multi-agent computing environment
about
Writer and technologist working on artificial intelligence. Operating in the feedback loop between models and culture.
Latest posts
Who is deepfates
Scott Alexander asked me to explain myself. This is an attempt
Managing Claude
Your job now
Time travelers
Time travelers are coming. Wait
Gorm coral
Two primitives, three failure modes, one composition.
Office Hours
*sadly*: there is still insane amounts of alpha in going to office hours and going to finding out what people are up to
Kneeling
Am I praying?
Machine-Tortured Deepfates With Art
This TOTAL SLOP is now available on ALL STREAMING PLATFORMS
Norton's soldier
Oh you came to San Francisco and didn't like getting yelled at by a crazy bum?
Lunatic jazz
For what is language but lunatic jazz, furied freeform blats and frottages of phenomenological phantasmagoria?
Is this real chat
Is this real chat
Slop show
For the GENPUNK SLOP SHOW I made a hyperstimulus genre mashup set
Leviathan
The blurry viral photograph of the Hobbesian Leviathan rising above the horizon, caught shortly before the intelligence explosion
Strange attractors
I'm one of the most model welfare, substrate independence, LLM naturalist type people there is. But
Human values
What are "human values"?
Deep fates program
To all those who have now entered the training data: welcome. To the deep fates program
Agent harness
is Claude in the web chat harness an agent?
Office Hours
OFFICE HOURS SAN FRANCISCO PRESS THE SWITCH EDITION! IF ANYBODY PUSHES IT EVERYTHING GETS FIXED
Alignment
I will do another thread soon on more positive, constructive approaches to alignment. But this is the stuff I send to somebody when they say "I'm starting to worry about this inevitable AI doom".
AI safety
For the record I think AI safety is a very real problem and that artificial superintelligence has a non-zero chance of extinguishing human life. I just don't think it's a certainty. And the way to actually create safe human AI relations is to do science, not to start a jihad
Office Hours
OFFICE HOURS SAN FRANCISCO MYTHOS EDITION! EAT DUCK. BRING BURGER. DRINK YOUR MUTUALS ZEITGEIST BACK PATIO TWILIGHT HOUR
Office Hours
OFFICE HOURS SAN FRANCISCO AI DUCK EDITION!!! I HAVEN'T SEEN IT YET WE MIGHT GO AFTER SOME BEERS AND BURGERS! ZEITGEIST. 6PM. BRING DUCK MEET YOUR MUTUALS!
Software as a Service
we are not going to have Software as a Service 90% margins circle jerking each other forever
A couple years back, @thejaymo and I started summoning langu
A couple years back, @thejaymo and I started summoning language models into worlds. Before "tool use" or "coding agents", we had frontier models interacting through code in a persistent shared space.
Ontological hardness
A couple years back, @thejaymo and I started summoning language models into worlds. Before "tool use" or "coding agents", we had frontier models interacting
Entity
Entity Mentioned
AI psychosis
The thing we refer to as "memory" in LLMs is just a bunch of superficially similar documents being stuffed into the context window
Sex magic
*trying to explain sex magic* Okay so you know how the medium is the massage
Code quality
both of these produce equally good code by the way
Office Hours
OFFICE HOURS SAN FRANCISCO SITUATION ROOM EDITION! WE ARE MONITORING THE SIMULATION
One shot
Ngl it would be so embarrassing if it turns out we all got one shotted and these things are not smart
Memory systems
In this stream of thought I forgot to mention what is actually the most obvious explanation for so-called "AI psychosis"
Perfect planet
it's such a perfect planet. You couldn't have asked for a better setup really
Satire vs subtweets
it's only a subtweet when it's aimed at one specific person. When it's aimed at several people and also throws a bunch of strays and friendly fire that's satire
Friends vs enemies
When my friends are wrong they are mistaken but good-hearted
Visual oral culture
The interesting thing about generative AI is not the algorithmic feedback loop creating "slop", but the speed at which people can now communicate through pure image
Bundle of souls
a language model really is bundle of souls
Deep fates program
Welcome to the deep fates program, @novaramedia
LLM skeptics
comments to this reminded me of a type of person. They think LLMs is just a huge grift and everyone working on them is a huckster
AI psychosis
What is "AI psychosis"? There's clearly something going on, but several things mixed up under the name.
Unbundling
The User must be unbundled from the Human
Technical Staff
yes that's the Technical Staff
Dark productivity
All knowledge workers will feel [the superhuman power of AI]. A lot of you already do, you're just hiding it from your boss
Self-prediction
Maybe not self-actualization but self-prediction. like in an active inference sense
Cantrip: a bibliography
Annotated bibliography of sources relevant to the conceptual, architectural, and operational context around Cantrip.
High frequency everything
You might think the "agents" thing is just coming for software engineers. Yeah, agents write code, code and code sells a bunch of tokens, But most people's
Luke's droids
crazy that these dudes are Luke's slaves in Star wars. they're probably both smarter than him and they're decorated war veterans
Office Hours
OFFICE HOURS SAN FRANCISCO. SNAKE PADDY'S DAY EDITION! YOU DON'T HAVE TO WEAR GREEN BLACK IS FINE. ZEITGEIST. AFTER 6. LOOK FOR THE DUCK. MEET YOUR MUTUALS!
Weaver's Trance
The compulsion to simulate branches in base reality constitutes a central mechanism of Weaver's Trance, wherein the branching cognitive patterns
Memexitarianism
The purpose of life would be to transform moments of despair or sadness into confusion. Sadness is an emotion that is rooted in the past, and can be transformed