Digital garden
Work in progress 🏗️Work of love
Garden contents
“What if we move from crafting static experiences to shaping how things behave, communicate, and participate inside intelligent, evolving ecosystems? We begin by designing with semantic clarity. Not just for aesthetics or usability, but for machine readability, intent encoding, and context awareness. That means: Structuring content with metadata, models, and meaning Creating systems that expose their logic, not just their appearance Designing not only for humans, but for the interpreters humans now rely on Treating every output as an input for another interpreter”
Grammar of intent

DesignPref: Capturing Personal Preferences in Visual Design Generation
“Current approaches to the development and deployment of AI agent systems do not support effective human oversight – they contribute to its degradation.”
AI Agents Push Humans Out of the Loop
“I think what’s going on in the United States now is that it’s all happening”
What Is the Perfect Color Worth?
“The history of labor shows that technology does not usually drive social change. On the contrary, social change is typically driven by decisions we make about how to organize our world. Only later does technology swoop in, accelerating and consolidating those changes.”
Opinion
“The very act of being an overseer degrades the capacities oversight requires: Oversight degrades the overseer.”
AI Agents Push Humans Out of the Loop
Put-That-There, 1982 - YouTube
Has Taking the Perfect Photo Ruined Tourism in “The Spectacle"? | The New Yorker Documentary - YouTube

Brian Roemmele
@BrianRoemmele
It was 1953 and Ray and Charles Eames created a masterpiece film for @BellLabs. Called “A Communications Primer” it introduces Claude Shannon and the concept of signal and noise as well as binary numbers. It has a beautiful “Shannon limit” introduction that still seems fresh. https://t.co/lM1aZxvtqt
December 20, 2019
“there can come a point when a corporation stops being just a company and begins seizing for itself the powers and trappings typically associated with a nation-state.”
Opinion
““A commercial company enslaved a nation comprising 200 million people.””
Opinion
“In the age of wooden television in the South where I grew up, leisure involved sitting on screened porches, smoking cigarettes, drinking iced tea, engaging in conversation and staring into space. It might also involve fishing. Sometimes the Web does remind me of fishing. It never reminds me of conversation, although it can feel a lot like staring into space.”
The Net Is a Waste of Time
“some think it remarkable that I do not use E-mail. In all truth, I have avoided it because I am lazy and enjoy staring blankly into space (which is also the space where novels come from) and because unanswered mail, E- or otherwise, is a source of discomfort.”
The Net Is a Waste of Time

citizenhush’s old account - RIP
@Hushisded
How selective sampling works https://t.co/kqqEwDUCsa
August 6, 2018
In Thai, when we eat something that's not to our taste, we don't really say "I don't like this" or "this is bad" @chezpim on X

Size of Life
“A bullshit job is one that even the person doing it secretly believes need not, or should not, exist.”
Bullshit jobs and the yoke of managerial feudalism
“The basic irony of standards is the simple fact that there is no standard way to create a standard, nor is there even a standard definition of “standard.””
Opinion
“a belief is a prediction, and learning is belief revision.”
Semantic Bayesian World Models
“We therefore argue that language models cannot scale to superhuman intelligence without organising knowledge in a semantic and probabilistic framework”
Semantic Bayesian World Models
“An autonomous agent must act before it knows.”
Semantic Bayesian World Models
“The privacy paradox observes that people often say they care very much about their online privacy, but willingly give it up in practice in exchange for a benefit”
Creating General User Models from Computer Use
“Despite progress in user modeling, recommender systems, and context-aware systems, computers remain remarkably unaware of who we are, what we are doing, and what would be helpful. At their core, current user models are simply too narrow: they understand our music preferences, our use of tools within a single app, or the TV show we might watch next. Even when user models integrate data across multiple applications, the integration remains surface-level; user models cannot reason or make inferences in new contexts. Our visions of technology, however, require user models that are broad, able to reason about everything from our general preferences to our immediate information needs; and capable of applying these insights across contexts, from work-related tasks to recreational activities. Applications today stumble because they have pinhole views of users”
Creating General User Models from Computer Use