The Mess of Things
On friction, efficiency, and what we give up when we optimize creative work
Since last September, I’ve been publishing two conversations a month on this Substack, and I want to reflect on what I have learned from the people I’ve had the pleasure of chatting with so far and revisit some of the core ideas that were the impetus for this endeavor in the first place.
One thing that has stood out across most of my conversations is the idea of slowness and intentionality as a point of importance for most artists, something I’ve begun to call friction. This is where the creative process demands something of the artist, either because of its own limitations or because of the underlying value that creative people place on a particular way of working. Friction in this sense is a good thing.
Anthropic recently released findings from interviews with 1,250 professionals about AI use, including 125 creatives. 97% of creatives reported that AI saves them time, and 68% said it improved their work quality. Although these numbers are interesting, even impressive in the right context, I wonder how they define quality, why saving time is important, and to whom. A separate survey by Metalabel Studios and Artist Corporations, which collected anonymous responses from 300+ artists and creators, suggests the view from inside the creative community is more ambivalent. 56% of respondents said AI will not change how they make creative work, not total rejection per se, but a deliberate compartmentalization.
I have two concerns here: One is this dichotomy between speed/efficiency and deep thinking/creativity on the other side. I suspect this is not the best way to frame the challenges of creative work. Automation and new workflow adoption are very useful in very particular areas, including rote tasks, preliminary research, and boring logistical work.
In my own practice, I offload certain kinds of labor that I don’t want to do to AI-assisted tools, but I want to keep the creative process as “inefficient” or as “slow” as possible. I put those words in air quotes only because their definitions and use are economic in nature, where speed = value. But for much of art making, value is generated in the doing, thinking, and being lost in the mess of things. Speed, in this context, holds little to no value with respect to deep understanding.
This brings me to the second point: I’ve found that where artists want help is often not where automation can provide it. I think I heard someone on the internet say that they don’t want AI to do their writing for them, but they’d love it if it did the dishes, folded the laundry, or called the telephone company to inquire about an unexpected charge. Although the sentiment was meant as a joke, I think there’s something to that.
This isn’t just about AI, exactly. It’s about what we give up when friction is engineered out of the embodied working experience. There’s a fair amount of research to support this kind of learning, too. Embodied cognition suggests that physical engagement creates memory pathways unavailable through passive observation or automated processes. When Dan Estabrook’s students coat paper with chemicals, for example, they’re not just making images; they’re encoding understanding through their bodies in ways that can’t be replicated by frictionless tools.
To go further, Maria Mavropoulou crystallized something I’d been wrestling with: “When I’m photographing, I’m not saying which camera model I used because it doesn’t matter. These are my decisions, but when you are working with AI, there’s you, the operator, and AI, which has its own agenda in the creative process.”
The boundary between human and machine agency is drifting and not clearly delineated. Photography faced similar questions at its inception; wasn’t the camera doing the work? But AI operates in the domain of decision-making itself. It’s not just capturing what exists; it’s generating what could exist, proposing alternatives you might not have considered.
The Anthropic data showed creatives described their AI use as 65% augmentative and 35% automative, but actual Claude conversations showed a more even split: 47% augmentation and 49% automation. This gap suggests we might be underestimating how much agency we’re ceding, or adapting outputs in ways not visible in conversation logs.
Travis LeRoy Southworth and I discussed this in terms of compressed time and labor. Work that once took days or weeks is now expected in minutes, not for deeper artistic reasons, but to satisfy the ever-growing needs of consumption. When you collapse that space, you collapse the opportunity for discovery. The accidents, the failures, the detours. These aren’t inefficiencies to optimize away, but the essence of creative processes.
When I use genAI, I actually don’t like the process. But I find value in arguing with the results, in the friction of disagreement. The act of struggling through the process is precisely where knowledge and creativity emerge. I, personally, try to create more opportunities to argue with my tools and understand their assumptions about knowledge formation in the process. I do not assume that a tool knows more than I do, even if they are faster in getting to a possible solution, because I recognize that these tools are optimized for a different kind of value production than my own.
I think there’s reason to believe that coming to certain technologies with the expectation that no matter how “good” or “efficient” they are, maybe artists aren’t striving for goodness or efficiency in the first place. I think that’s why I’ve really enjoyed talking to creative folks who straddle the line between commercial and fine art practices, that even if the outputs seem indistinguishable to their audience, they are different modes of relation with distinctly unique concerns that underpin them.
In my conversations, I keep asking: What are you trying to protect? What aspects of your practice are non-negotiable? For some, it’s the research process or serendipitous discoveries that emerge from diving deep. For others, it’s material engagement, be it the feel of the medium or the physical labor of making. And for others, it’s decision-making authority.
Like any healthy relationship, it seems like the people most at peace with technology use have articulated clear boundaries, even knowing those boundaries might shift. They’ve identified what’s at stake for them personally and made deliberate choices about where to draw the line, for now.
Many of you who are reading this are creative people, so perhaps these things are not that revelatory. But I keep circling them, and think what I’m actually seeing and struggling with is a conflation of different aspects of work that are both creative in nature, require some amount of labor, but ultimately have different expectations attached to them.
There is the work we do to survive, and there is the work we do to live.
We have entered an era where the possibility of supporting ourselves through work that is both surviving and living seems on, it’s surface, plausible. Artists are able to sell their work on marketplaces online without gallery oversight, build communities online and elsewhere to support grassroots initiatives, and find editorial and commercial clients more receptive to hiring artists with their own unique voice. There is a merging of industries that has many incredible affordances, as photographer Carlo Van de Roer shared in our conversation, but also some pressures worth naming. The ACF survey puts it starkly: the median annual wage for U.S. arts and design workers is $52,000. The median annual rent for a one-bedroom apartment in New York City is $45,000. Nearly half of the respondents expect to feel less financially secure as a creative person in 2026.
That paradox from the Anthropic data makes complete sense in this context. Artists are satisfied because these tools genuinely help them work more efficiently, explore new territories, and solve technical problems. But efficiency is most appealing when you’re underwater, and we should be careful not to mistake it for value in the process.
Mavropoulou said, “Art is about translating your own experiences through a form and then communicating that to your viewers. Art is made by humans to be seen by other humans. Of course, we might use tools to do that, but it’s still a conversation between humans.”
When we collapse the friction that forces us to translate our experiences ourselves, we risk losing the human specificity that makes work resonate. Not because AI, or any other technology, can’t generate compelling images, texts, or ideas but because the process of wrestling with material, with tools, and most importantly, with our own intentions, that’s where the ah-ha moments happen.