Responsibility survives at every step in quantities too small for anyone to possess it.

This particular sermon is about automation, but not really the familiar problem of a machine being allowed to make important decisions all by itself. Its more unsettling (and common) case is a system in which plenty of humans remain very much involved. The machine recommends. A person reviews. Another approves. Another implements. There are policies, thresholds, audit logs…and yet things can happen that none of them quite decided.

There is a comforting phrase for the systems designed to prevent this: human in the loop. Whatever the software is doing, a person remains somewhere in the process, able to exercise judgment before the decision becomes a real thing. 

Between 2013 and 2015, the State of Michigan’s MiDAS unemployment system automatically identified people it believed had committed unemployment fraud and, for a period, could resolve those cases without human intervention. The state eventually went back through tens of thousands of cases. Out of more than 40,000 fraud findings originally resolved by computer, 85% were reversed.

Well, this is why a human belongs in the loop, right?

Well... Michigan also reviewed another 22K cases that had been initiated by the computer and then referred to an investigator. 44% of those fraud findings were reversed too.

These were different sets of cases, so we can’t say how much the investigators helped. But their involvement didn’t prevent thousands of fraud findings that the state later reversed.

By the time a person sits down in front of their APPROVE button (or their Exec-u-Calm™ Decision Maker) , a lot has already happened. A system decided which information is relevant. A model has assigned a score. Somebody determined the threshold at which that score becomes suspicious. A queue placed this case ahead of that one. The UI chose what to prioritize and what to bury. A policy has established which actions are considered “normal.” Then we point to the domino at the end of the chain and say: all good, there’s a human in the loop.

Can this person understand where the recommendation came from? Do they have enough time to disagree? Is disagreement operationally realistic? It might technically be possible to reject something, but if I’m working inside a process that makes rejection unusual, expensive, slow, or professionally risky, it gets tricky.

Obviously, decisions are distributed on purpose all the time. Surgeons don’t manufacture drugs, maintain imaging machines, establish hospital safety policies, or personally verify every instrument before operations. Aviation is full of overlapping procedures, automated systems, and people responsible for different pieces of the same flight.

So the problem isn’t that nobody possesses the entire decision. The issue is whether somebody is responsible for how the pieces behave together.

If hundreds of employees repeatedly approve recommendations from a system, somebody needs to notice what the pattern of approvals, overrides, appeals, and reversals says about the machinery producing them. The exception cannot always be treated as an isolated failure by the person who happened to touch the case last.

A 2020 UK government review of algorithmic decision-making makes a distinction between humans in the loop and humans over the loop: people who understand the system well enough to monitor the fairness and effectiveness of the whole decision process and who carry responsibility for it.

A person over the loop is able to reconstruct how choices become possible. Who chose these inputs? Who set this threshold? Who decided what our reviewers would see?

The person in the loop encounters a case. The person over the loop needs to encounter the system, to see what the chain of dominos is doing.

To bring it over to design for a minute, the idea of designing systems rather than iterating on artifacts is not new. Grid systems, style sheets, brand guidelines: we’re used to thinking about how individual choices add up to something larger.

If approving a recommendation takes one click, but questioning it means digging through supporting information and writing a justification, we’ve made agreement easier than disagreement. Requiring a justification can be reasonable. But if the extra work discourages reviewers from raising valid objections, we risk building a dark pattern. These are crucial design decisions about how judgment gets exercised, and we’ve made them before a reviewer ever sees a case.

Which means our responsibility has to extend beyond the click. We need to sit shotgun while people use the thing. See whether they can or will challenge it. And see whether those challenges can reach anyone with the authority to do something about it.

Read “The Workflow Is an Alibi”

It sounds trivial written out, but there is a real change in attention hidden inside it. Instead of treating the landscape as empty space separating me from destinations, I start treating it as a field of possible interactions.

The adventure-game metaphor helps because it changes what counts as significant. A strange staircase becomes a path. A shop becomes an encounter. A distant structure becomes something worth moving toward, at least until something else catches my attention. A dead end isn’t wasted time; it is explored territory. The whole thing gives me permission to be inefficient.

The basic rule is simple: treat everything as potentially interesting, and treat public access as a constraint of the game.

That constraint turns out to create exploration. A fenced property isn’t something I cross; it becomes terrain I have to route around. A river without a bridge becomes a world boundary. A footpath becomes valuable because it connects two regions that otherwise seem separate. Public transportation becomes fast travel.

Moving around this way starts to reveal a different topology of where I live: not the version organized around the fastest route from one known destination to another, but the one made out of paths, barriers, crossings, shortcuts, edges, invitations, and things I would normally pass without noticing.

• Reading “Strange Houses” nightly with my son. Such an inventive way to tell a creepy story, and that's all I'll say because it's unusually easy to spoil. Its author, “Uketsu”, is a masked Japanese YouTuber who originally told the story as a 21-minute mystery built around floor plans before it became a novel. A book with some evidence-folder DNA.
• Looking for my next job, using an AI workflow I've been developing to help with the search. It's become an interesting project in its own right. I'll write up the process once I've been through enough of it to know what actually works.
• Enjoying summer with the kids. Lots of hiking, biking, getting sweaty, getting lost, getting unlost, and generally remembering that going outside remains pretty competitive technology.
• Building a catio behind the house. A “catio” is basically a screened outdoor room for indoor cats, a little loophole where they can smell the grass, lie in the sun, and contemplate birds without terrorizing the local ecosystem.

I’m drawn to the almost embarrassing obviousness of it. People can’t see the bus. Take the stickers off. It pokes at something I’ve become increasingly suspicious of in design, especially design thinking: we can get so interested in demonstrating that thoughtful design has happened that the evidence of our thoughtfulness starts competing with the thing itself. The research, the workshop, the framework, the strategy, the explanation of why every single decision is meaningful. I’m not against those things; I use them for sure. But sometimes they become a very expensive way of noticing that window with the stickers.

Then I catch myself turning the bus station into a little morality play. Quiet, useful design is good. Visible, expressive design is vain. That would flatter my own taste pretty efficiently. Morrison doesn’t quite stay there either. He goes on to describe successful objects as neither purely aesthetic nor purely functional, but shaped by materials, their surroundings, communication, and the experience of living with them. What got me, I think, isn’t utility alone but the way value appears partly outside the thing, in what changes around it.

The version of functionalism I’m wary of—form serving a named purpose—gets me part of the way. The café windows have a job and do it much better once the stickers are off. But I’m interested in a looser question: what does the design make possible? In that model, the function is named in advance. Possibility may only show up in the encounter, and it may belong as much to the person using the thing as to the person who designed it. Sometimes it’s practical. Sometimes it’s awe, pleasure, curiosity, attention, or skill. I don’t want to smuggle all of those back into function and pretend the distinction has disappeared.

Kenya Hara’s comparison of two kitchen knives gave me a way to feel that difference rather than define it. A German Henckels knife has a fluid, ergonomic handle; the thumb finds its place. The Japanese knife has a flat, unshaped handle that leaves more room for a cook’s skill. Hara is describing different flavors of simplicity. Both can be generous. What becomes possible isn’t always what becomes easier.

It also keeps me from using Morrison as an alibi for minimalism, which would be convenient because restrained design is already close to my own taste. Minimalism can become just as narcissistic as ornament; a very quiet object can spend all day long telling me how exquisitely quiet it is.

Visible thoughtfulness can be generous. There are objects I love partly because I can see that somebody obsessed over them: a beautifully drawn letter, a joint made far better than necessary, an animation somebody clearly spent too long getting right. Sometimes the evidence of effort, the delight in the craft, is part of what I’m receiving. It tells me somebody cared enough to keep going after the sensible stopping point.

I keep thinking of this as “extravagant commitment”: somebody deciding that a thing deserves far more effort than pure efficiency can justify. Religious architecture makes that hard to ignore. The Sagrada Família doesn’t politely disappear into use. Its height, light, ornament, material, and accumulated labor are part of what you encounter. It’s trying to alter your state. I could call awe a function, but at that point function starts to become one of those suitcases you have to sit on to get the zipper closed.

The 10,000-year clock does something similar in a completely different register. It’s an immense mechanical clock being assembled inside a mountain, intended to keep time for ten millennia. Nobody needed to build that. That’s partly why I find it great. The overkill is part of the thought: ten thousand years stops being a number and becomes a machine somebody is actually trying to make work very, very, very well.

None of this makes my suspicion of design narcissism disappear; it just moves the line. I don’t think the problem is that the designer is visible. Sometimes seeing the hand is the gift, and “look at this” can be generous. I get suspicious when the work folds back toward the designer—when the main experience on offer is admiration for the intelligence, taste, or sensitivity of the person who made it. Designers can spend a lot of time making things for other designers, where the sophistication is easiest to appreciate if you already know the references, the vocabulary, and approximately how much the typeface cost.

Even that doesn’t give me a clean rule. Someone can make something out of vanity and accidentally give people enormous pleasure. Someone else can approach a problem with genuine empathy, a bazillion user interviews, and an immaculate workshop and still make something dead. Intent helps me understand the work, but it doesn’t get the final vote. Neither do restraint, ornament, craft, or usefulness.

This is also how I think about people. Some people are impressive as objects of attention; others change what becomes possible around them. Around them, I become funnier, more curious, less guarded. I say the unfinished thing. I try something I might otherwise have talked myself out of. I don’t mean this as a morality test—a charismatic asshole can unlock plenty. It just reminds me that value isn’t always contained inside the person or object. Sometimes it appears in the relationship.

Back at Fisch Platz, the result is almost too ordinary to photograph: someone sits in the café, sees the bus arrive, and gets up. I don’t want every design to disappear that completely; sometimes I want it to stop me in my tracks. Either way, I’m interested in what happens next.

The site runs deep. Games, art projects, essays, software, photographs, talks, science, technical standards, jokes, personal histories, and a library database containing thousands of books are all in there. His so-called “redundant site map” is less a navigation aid than a warning about what you’re getting yourself into.

Twenty years is a long time to rummage around in somebody else’s basement. I’ve watched projects appear, followed his game development, read essays about things I barely understood, and periodically fallen into parts of the site I’d somehow missed the previous twelve times through.

His book list alone has sent me toward a ton of great fantasy and science-fiction books I never would have found otherwise. It’s his actual library database: more than three thousand books at this point, with years of comments scattered through it. That’s part of what makes it useful to me. I’m not reading Andrew Plotkin’s Ten Essential Fantasy Novels. I’m snooping around the bookshelves of someone whose taste resonates with mine.

His games are the reason I arrived, and they remain extraordinary. His interactive-fiction page runs from early experiments through The Dreamhold, Shade, Spider and Web, and a pile of stranger little things. Then there’s Hadean Lands, his enormous alchemical puzzle box and, for me, his magnum opus. The site doesn’t arrange this work into a grand career arc. It reads more like: this is what I’ve been doing.

The standard professional website compresses a person. Here are the three things I am good at. Here are the six projects that prove it. Here is the sharpened paragraph where I explain my unique point of view.

His site does the opposite. The About Zarfhome page says plainly that the site is about what he does and how he does it. He talks about showing drafts, tools, half-baked ideas, abandoned ideas, finished things, and failures. His phrase for it is “work in daylight.” He makes a distinction I’ve carried around for years: being a professional at one thing doesn’t mean the rest of what you make has to disappear.

The site is mostly plain HTML, and the About page tells the reader to “choose your own damn default background color.” The plainness isn’t neglect. It’s a decision to make the site durable, to let it work anywhere rather than control every pixel.

A person keeps putting things somewhere for long enough and eventually patterns emerge that no About paragraph could have manufactured. The kinds of problems he returns to. Books he buys. Jokes he repeats. Technical standards and systems he helps maintain. The game that took years longer than expected. The thing he was excited about in 1997 and the thing he’s still thinking about in 2025.

Because enough of the material survives, the portrait isn’t clean. Even the About page contains a little crack in its own philosophy. In 2000, Plotkin announces his desire to show unfinished projects and failures. In the 2010 update, he admits that, well, he hasn’t done as well at that as he wanted to. As his standards rose, fewer unfinished things made it onto the site. He labels the problem vanity. So the guy who built a website partly to resist hiding unfinished work eventually became self-conscious enough to start hiding more unfinished work.

The site doesn’t give me access to him in some pure, unfiltered sense. I don’t know him. He’s choosing what goes on the site just like anybody else chooses what to show.

But after twenty years, those choices have left a pattern: what he keeps, what he returns to, what he revises, what he admits he withheld. There’s enough material there that I can feel the contours of somebody’s attention.

Dru Riley’s “100 Rules to Live By” feels like the residue of this process. He says each rule points back to a personal story. Somewhere behind Dru’s “Don’t self-reject” or “Make it work. THEN make it pretty” is an incident I’ll never know. The incident has disappeared, but its shape remains. That feels a lot like folklore: experience getting worn smooth through repetition until what survives is the part somebody decided was worth carrying forward.

A family has three disastrous camping trips and ten years later everyone says, “We’re not camping people.” Nobody remembers the exact severity rainstorm, the reasons for the broken tent, or who was miserable first. Incidents blur together and the verdicts survive. I’m doing this with myself all the time. One episode comes to stand for ten others. I forget where a conviction came from but keep carrying it around. Past me leaves behind warnings, superstitions, shortcuts, little scraps of alleged wisdom for whoever I become next.

Some of them are useful and some are assuredly complete nonsense, which is what makes the whole thing interesting. Folklore isn’t truth. “I always do this” is folklore. “I’m not the kind of person who…” is folklore too. A story can begin as a pretty accurate description and quietly harden into identity. Three bad camping trips become a family trait.

This makes a journal seem less like a record of a stable self and more like a place where a mythology is being manufactured in real time. I’m remembering, simplifying, exaggerating, finding patterns that are probably real, and occasionally drawing a constellation between completely unrelated stars. Years later, I inherit the whole mess from the people I used to be: useful warnings alongside old misunderstandings, good instincts alongside conclusions that may have expired without telling me.

Riley’s list feels like a lifetime of personal folklore compressed down to a portable size. It's a great list, and it reminds me to be aware of which stories I keep telling myself, where they came from, and whether they’re still true.

Sartre was absolutely electrified by the exchange. He got himself over to Berlin to study this stuff for himself, and voilà! Phenomenology becomes a major foundation on which his own philosophy was built.

I love this story because it humiliates philosophy juuuust enough. Aron doesn’t begin with Being, or Truth, or the Soul. He points at booze on the table. It’s easy to hear this as a promise to make philosophy more accessible, as though phenomenology were saying: oh don’t worry, we can use relatable examples. But that’s not really what’s interesting about it to me. The encounter with the drink is where the philosophy begins, which is a much stranger proposition.

I have a habit of assuming that serious thought lives at some higher altitude. If we want to think about something deeply enough, we must ascend out of the café and start looking for those inscrutable forces bearing capital letters: Meaning, Consciousness, Freedom, Art, God, Reality. Meanwhile, the glass is sitting there, sweating onto the table.

One thing phenomenology says is to pay very close attention to experience as it’s actually lived. Not merely hey, that’s an object over there, but the experience of perceiving it, remembering it, wanting it, recoiling from it, recognizing it. Husserl gets considerably more complicated than staring down drinks, obviously. But the concept that smacked Sartre so hard in this encounter was that experience itself could be philosophically inexhaustible.

What exactly is this cocktail before I file it away under “cocktail?” Well…there’s color, which I recognize as apricot even though…hmm…I’m not entirely sure apricot is a color. There’s the cold glass in my hand. Sweetness arriving before bitterness. My expectation of what it will taste like. The café around it. The fact that I’m sitting with these particular people. One of them is watching me drink it, etc. All these memories and preferences and tiny private histories that have somehow become involved in this otherwise innocent beverage.

Naming things is useful, but it also shrink-wraps them. Cocktail. Café. Friend. Stranger. Job. Website. Tree. Myself. Once something has a name, it becomes very easy to stop encountering it. We know what that is. Move along. The world becomes incredibly efficient and navigable this way, and also slightly dead.

To me, there are actually two encounters happening at that table. Sartre is encountering the cocktail differently, but he’s also encountering Aron: somebody who had gone somewhere, found something that rearranged the furniture in his head, and come back excited enough to try to make his friends see what he’d seen.

Aron didn’t hand Sartre a book or paper, or give him an academic lecture. He looked around the table, found the nearest available object, and said, essentially: this.

There’s something I recognize in that. Some of the most important things I’ve learned from people have arrived badly packaged. A friend gets completely obsessed with a book and starts explaining it before he quite knows how. Someone sends a link saying OMGGGG!!!!!!! You visit somebody’s house and notice the fascinating little system they’ve developed for doing something you’d never previously considered a system at all. Another person has been paying attention somewhere you haven’t, and for a moment you get to borrow their eyes.

The idea totally matters, but so does the enthusiasm. There’s a kind of intellectual friendship in which people don’t simply exchange conclusions…they infect one another with attention. You become interested because somebody you trust has become interested, and their interest is specific enough that you can get utterly electrified by proxy.

I think this is a piece of why Bakewell’s account of existentialism starts here. Existentialism wasn’t produced by a bunch of isolated genius brains emitting streams of books. It developed in a nest friendships, love affairs, arguments, rivalries, cafés, letters, teachers, students, and people enthusiastically mangling one another’s ideas.

Origin stories are suspiciously tidy, of course. Intellectual movements aren’t born on a table one afternoon any more than rivers begin beneath signs saying RIVER STARTS HERE. Sartre already had his own philosophical ambitions. Husserl had spent decades developing phenomenology. Other thinkers were shooting those ideas all over Europe. I think the legend survives because it compresses something messier into a scene you can hold.

The exchange over the drink seems almost embarrassingly small compared with what followed: Sartre, Beauvoir, existentialism, Being and Nothingness, decades of argument about freedom, consciousness, responsibility, and what the hell we’re all doing here. But that’s exactly what I like about it. The big thing doesn’t arrive as a big thing. It arrives as your friend leaning across a café table and saying: look at this.

Kris Sowersby removed some of the more noticeable details, lost the thread, put the project away, returned to it years later, and nearly threw it out. It was only after reading about Super Normal that he found a way to understand what he’d been making: maybe being deliberately ordinary was enough of a reason for a typeface to exist.

I love that the explanation arrived late. That feels closer to how making usually works: the work moves before the story catches up.

Sowersby describes deliberately removing traces of his own recognizable style from the letters, which sounds almost perverse to my designer brain. I'm encouraged to make things distinctive, ownable, memorable—something that can stand up in a presentation and explain why it deserves to exist.

But most of the time, when I’m reading, I don’t need the typeface to introduce itself. I need it to help me read.

That’s probably why Untitled Sans feels right for both my portfolio and for thinking.haus. The work and the writing can wander around and follow unusual ideas. The letters stay put.

Its plainness isn’t an absence of design. It’s what happens when a mountain of small decisions are made thoughtfully enough that they can recede. The type makes room around the words instead of elbowing them for attention.

I find that deeply appealing: work that has clearly been thought about, but doesn’t keep asking you to admire how much thinking went into it. Work that knows how to be quiet.

Read Kris Sowersby’s notes on the Untitled Collection

He says, “In the beginner’s mind there are many possibilities, but in the expert’s there are few.”

Expertise is darn useful. It recognizes patterns, avoids old mistakes, knows which details matter. But it can also become a kind of compression. The more I know about something, the easier it is to see the category before I see the thing itself.

Suzuki isn’t asking us to become stupid. Beginner’s mind isn’t the absence of knowledge or experience. It’s a way of standing inside what you know without letting your shoes turn to cement. I find this much harder than it sounds.

The book is assembled from talks and still feels spoken. Ideas appear, disappear, return slightly changed. Sometimes an explanation leaves me less certain than I was when it began, which seems appropriate.

Suzuki talks about posture, breathing, repetition, effort, and the very ordinary difficulty of sitting still. The spiritual life he describes is physical. It’s in my hands, my spine, my breath, and my willingness to sit there without turning the whole thing into another achievement.

This is a problem for me because I like practices that produce results. I want to know if I’m progressing. I want a little dashboard somewhere telling me my Zen is up 14 percent from last quarter. Even when I’m actively trying to become less ambitious, I can become ambitious about becoming less ambitious.

Suzuki keeps returning to the activity itself. Sit because you’re sitting. Breathe because you’re breathing. Do the thing without constantly climbing outside it to ask what it means, whether it’s working, and what all of this says about you.

I can’t say I consistently live that way. But I’ve come to value the freedom in it, and meditation gives me a place to practice it. Sometimes I notice how quickly I’ve decided who another person is, what something means, who’s right, who’s ridiculous. Occasionally I can catch myself before the judgment hardens and look again.

That may be what beginner’s mind means to me now. Not becoming innocent again, and certainly not forgetting what I’ve learned. Just setting it down for a minute. Loosening my grip long enough for something I wasn’t expecting to get in.

The joke, of course, is that I can immediately turn this into another thing to master. A program for becoming exceptionally good at having no ambitions about becoming exceptionally good.

Zen Mind, Beginner’s Mind doesn’t make Zen easy. It makes simplicity seem like enough to work on.

Read Zen Mind, Beginner’s Mind

A portfolio case study is an unfairly antiseptic account of a project. Here is a report. Here is an identity, a platform, a system. Everything appears as a finished object.

What disappears are the personalities, budgets, old decisions, new anxieties, unclear language, and deadlines that don’t care whether the idea is ready. For a long time, I thought of all of this as the stuff surrounding the work. The real work was what I could see. That got harder to believe as I started leading more projects.

Much of the invisible part is translation. Someone says they want “more impact.” Maybe they mean contrast. Maybe confidence. Maybe they’re worried the CEO is going to flip past the page without noticing it. The request and the actual worry are often two different animals, and asking one more question can often keep a room full of people from spending a week solving the wrong problem in excruciating detail.

There are days when I leave work having “made” almost nothing. Maybe I've clarified a decision, connected two people, moved something out of somebody’s way, given a designer another afternoon before showing the work, or written down enough context so we didn’t have go on an archaeological dig through Slack messages.

At first, those days made me vaguely guilty. I like making things. I like the haptic feedback of an idea taking shape in my hands. “I helped several conversations go better” has considerably less portfolio appeal. But I’ve come to appreciate that without nurturing the conditions for good work, no good work will happen.

You can see the symptoms of this hidden work: fewer surprises. Calmer handoffs. Decisions that still make sense a week later. Somebody else being able to pick up the thing without needing you standing beside them.

I still prefer the days when I make something I can see. But I no longer think those are the only days that count.

I used to confuse looking with collecting. I’d open a browser, save fifty images, and call the result research. My collection grew; my understanding did not.

Now I try to begin with a question. What am I actually trying to understand? What has become so familiar that no one sees it anymore? What does a person need to know, feel, or do—and in what order? The question doesn’t need to point toward an answer. It only needs to give my looking a shape.

I start near the subject. If I’m working on a corporate report, I look at reports. I need to understand the form before I decide what to do with it.

Then I move farther away. I might look at newspapers, maps, instruction manuals, or paintings. A newspaper knows how to handle density. A map knows how to hold several scales at once. A painting may know something useful about directing the eye.

I’m not looking for a style to borrow. I’m trying to notice what something is doing. When I save a reference, I write down why.

The note is at least as useful as the image. It gives me enough distance to carry the idea somewhere else without copying its appearance.

Then I arrange what I’ve found and start removing things. One reference tells me what I like. Several references placed together begin to reveal what I think. The one that doesn’t fit is often worth keeping around. Or it may be pointing to something the rest of the collection has politely avoided.

Time helps too. I’m easily persuaded by novelty and veneer. If I leave the references alone for a day, some become less convincing; a few become more interesting.

Eventually, I make something, usually before I feel ready. A rough sketch shows whether an observation can survive contact with actual words, scale, and use. If it can’t, no matter how cute it is, it gets tossed off the cliff.

So my system isn’t very complicated: begin with a question, look outside the category, name what matters, arrange, remove, make, and look again. The point is not to build an impressive moodboard. It’s to notice enough that a decision becomes easier.