I built blnkfld as a simple tool to make complex projects legible. Use it on any GitHub project to gain a footing and understand it better.
As agents write more of our software, I keep coming back to a question: how do we understand what they’ve built? I want to see how the pieces connect, follow an action through the system, and know where to look when working within it.
blnkfld turns a repository into an interactive, agent-generated brief. It gives you an overview of the project, a map of its architecture, and a trace of how one action moves through the code. The explanations point back to the source, so you can keep exploring.
Run it from your repository, from any folder:
npx blnkfldI think there’s room for the README to become something more explorable; a starting point for asking better questions about the software in front of you.
blnkfld is ready to use and still taking shape.
Take the interface away and those old signals evaporate, while the stakes explode at the same time, because this new layer reads your mail, spends your money, and speaks in your name. It is no accident that the loudest conversation around every product in this new category is about trust, about what it can access, what it stores, and what it might do when you are not watching. The biggest design problem in this category is no longer usability. It is legitimacy.
So how do you design for trust when there is no interface to carry it? I believe it lives in three places. It lives in how the agent communicates, in its tone, its honesty, and its restraint, in whether it admits what it doesn’t know and asks before it assumes, because every sentence is now a design decision and the conversation itself becomes the brand. It lives in how you show what is happening on the other side, not through a wall of logs but through a considered kind of transparency, the reassuring feeling that the machine’s work is visible whenever you care to look. And it lives in the approval moment, that brief flash where an interface does appear right before an action is taken, carrying exactly enough context for a confident yes or no. As the interface shrinks, that moment becomes one of the highest-craft surfaces in all of software, because the fewer pixels that remain, the more each one weighs.
Design isn’t disappearing along with the interface. It is concentrating, and all the care we used to spread across hundreds of screens now pours into a handful of moments, into a sentence, a summary, a request for permission. Those moments will decide which services people let into their lives and which ones they quietly disconnect.
As our world continues to change rapidly and become more complex, systems thinking will help us manage, adapt, and see the wide range of choices we have before us. It is a way of thinking that gives us the freedom to identify root causes of problems and see new opportunities,.
So what is a system? A system is a set of things—people, cells, molecules, or whatever—interconnected in such a way that they produce their own pattern of behavior over time. The system may be buffeted, constricted, triggered, or driven by outside forces. But the system’s response to these forces is characteristic of itself, and that response is seldom simple in the real world.
SOURCE: THINKING IN SYSTEMS BY DONELLA H. MEADOWS, 1993.
While invisibility seems inevitable, things seem to constantly return to the fundamental importance of systems built on constraints which assume degradation as a byproduct of compounding growth.
The goal has never been to make every product identical. Rather, our goal was to establish shared expectations around common experiences. There’s no reason why every portal should have a different create flow. People moving between products should encounter familiar approaches to tables, detail pages, navigation models, and other recurring patterns.
Designers and engineers should start from common foundations rather than reinventing solutions repeatedly. Not only does this simplicity improve the customer’s experience, it also saves unnecessary production toil spent reimagining patterns over and over again in each portal experience.
Consistency—or lack thereof—was the symptom of a problem that simplicity could cure. ¹
In production I’d made an early investment on shadcn/ui as a library, having gone through the cycles of dabbling in MUI, Catalyst, among a few others. Being before the advent of hardcore agentic coding, it was a reasonable bet that luckily won over time, likely to both its community and its open-sourced nature ².
How things compound is a fundamental concern that is observed in both software engineering and design, or the intermingling of both. I’ve come to recently admire the living design lab that is Teenage Engineering, who’s core philosophy revolves around the notion that engineering is design, and proper guardrails and the idea of making products feel finite. This structure is what lends itself to this notion of “simplicity,” and I’ve learned in a particular pastime, the ingredients to utmost simplicity (and as a symptom, beauty) are likely, proper restraints.
Sentient Design applies perception, interpretation, and agency to deliver intelligent interfaces that are:
- Aware of user context and intent
- Radically adaptive across content, structure, behavior, and medium
- Emergent and conceived in real time
- Collaborative in shaping the experience with the user
- Individualized to moment and context, not to stale profiles
- Deferential to human judgment, especially in consequential moments
- Continuous across interaction modes
SOURCE: SENTIENT DESIGN, JOSH CLARK & VERONIKA KINDRED, 2026.
Machine learning
- Recommendation
- Prediction
- Classification
- Clustering
- Generation
Domains
Computer vision / Natural language / Speech
Knowledge-based
- Expert systems
- Knowledge graphs
Traditional Algorithms
- Search
- Sort
- Other logic
Fig. 02 — Three approaches to machine intelligence.
Traditional software and algorithms are deterministic: they follow fixed rules and logic that are predictable and repeatable. Machine learning systems are probabilistic: they work with likelihoods, not certainties. Their answers emerge from patterns and probabilities, not fixed logic or facts. This makes them remarkably flexible, but you can't design for exact outcomes.
What makes machine intelligence so powerful is how these three approaches work together. Machine learning brings adaptability and insight, knowledge-based systems supply grounded info, and traditional algorithms provide reliable, repeatable results. Just like Roman concrete, the right blend makes the whole stronger.
Machine intelligence promises to be today's concrete. Comparing AI to a building material might sound like we've inhaled too much dust, but like Roman cement, it has the potential to support experiences the world hasn't seen before if it's understood correctly.
Digital design materials like images, prose, code, and data may not be physical, but they still have their own grain and texture. HTML and CSS bend in certain ways but not others, while data demands density in some contexts and breathing room elsewhere.
Machine intelligence is the new design material in your kit. Just like you sprinkle JavaScript into a web page for interaction, or apply responsive design for device flexibility, you can now weave intelligence into the fabric of your experiences, too.
SOURCE: SENTIENT DESIGN, JOSH CLARK & VERONIKA KINDRED, 2026.
In the 1960s, the American anthropologist Edward T. Hall explained what is known as the Cathedral Effect. The idea that a high ceiling evokes feelings of freedom and openness, fostering abstract thinking, creativity, and holistic problem-solving in individuals. In contrast, low ceilings create a sense of confinement, encouraging focused, detail-orientated, and analytical thinking. ¹
I live in New York City. While not conducive towards solitude or focus in public spaces, a small ritual has proven effective for me: reading books on my phone every morning, commuting on the N/Q/R downtown. A low, noisy, and sometimes dimly lit subway carriage has become my daily respite for learning, a predictable block of focus for 10 minutes, underneath the city.
The purpose of this site, or state of this state moreover, is like a drawer cracked partially open, jumbled full of disparate objects of shape, color, and purpose representing thoughts. After months of mulling about in my head over this site, I’ve decided that it's important to imperfectly capture the small, scavenged artifacts from the world that I can hopefully trace back to from a not too distant future.
[1] How Ceiling Height Influences Creativity and Focus by Patricia Fierro-Newton, 2024