Does It Matter If You Used AI To Make It?
It depends
Thank you for being here. Please consider supporting my work by sharing my writing with a friend or taking out a paid subscription.
Last week a colleague shared a recent Kyle Chayka essay from the New Yorker titled The A.I.-Design Aesthetic That’s Taking Over the Internet. In it, Chayka chronicles a new phenomenon that many of you are likely becoming familiar with. It’s that feeling that you just can’t quite shake that whatever is in front of you smells like it was made with AI.
For many, it’s not just a smell, it’s a stench.
Chayka opens with an anecdote to set the scene. The archetype will be familiar to many of you, especially those of you who are terminally online.
Matt Ström-Awn, an independent designer, works mostly with startups in their early stages. Recently, two different clients proudly showed him sales decks that they had produced to court customers. Both decks had a bright-colored first slide with a mission statement, written in three declarative bullet points. Both had second slides featuring four rectangles laying out “the playing field”—the market in which the startup was operating—and both had third slides with a centered line of text reading “our move” and describing the startup’s disruptive tactic. “They actually look like they were generated by the same company,” Ström-Awn told me. “The logo is different, but the design is the same.” The decks looked the same because both were made using Claude Design, an A.I. tool that Anthropic launched in April. The new technology, Ström-Awn said, “defaults to the same aesthetic for every single person that’s using it.”
Chayka mentions a few examples in his piece to illustrate but doesn’t hyperlink to the sites. I did a little digging to pull them up so you can take a look. As you look at these sites, you can get a sense of the tells that the designer in Chayka’s story was latching on to.
Even if you haven’t looked at enough AI-generated visual artifacts (especially as seen in AI-generated websites and slide decks) to develop this sort of gut-level intuition when you see something, you’ve likely seen it in the AI-generated text that is increasingly flooding the web. From overzealous em dashed parentheticals to overused words like “delve” to phrases flooding our social feeds in the form of “is not X, but Y,” the tells are becoming quite obvious at this point. Once you see it, you can’t unsee it.
The thing itself is valuable
I appreciated this article because it raised a tension that I think we need to talk more about. Here’s one way to phrase it: “Does it matter if AI was used to make it?”
Of course, as in most situations, the answer is “it depends.”
I understand the disdain for these new AI-fueled design norms that are sweeping across the Internet. The Internet is becoming, and will increasingly become, a place that smells like AI. As AI continues to develop and be integrated into the digital tools that we’ve grown accustomed to using in our work—did you notice that the Gemini chat box is directly at the bottom of every new Google Doc you open now, hiding behind a subtle little bar with a sparkle on it?—we’ll inevitably need to answer this question for almost everything we do on a computer.
Does it matter if AI generates responses to my emails?
Does it matter if AI categorizes them?
What about AI generating a grant report from a set of data?
What about using AI to create a program to more effectively learn my students’ names?
In some instances the answer will be quite obvious, but in most cases there will be a lot of subtlety. If you are tempted to think that most situations are black and white, I’d push you to look a little more closely.
This tension has been exposed most clearly for me when I consider the things that I have created in the past year that would have never existed unless I used AI to help me build them, everything from websites to little software tools to AI agents that help me to more easily manage and organize my life,.In constraint-free scenarios where there are no limitations on the available expertise and sufficient time and money to hire those who have it, then perhaps there is an apples-to-apples comparison to be made. But the realities of the real world are quite different. There are many ways to solve a problem if you ignore one or more of the constraints.
Let me illustrate by way of example. As many of you know, earlier this year I helped to co-organize a conference at Harvey Mudd with a group of my colleagues from Mudd and across the Claremont Colleges. As we began to prepare for the conference, we were trying to figure out how we would communicate the schedule and sessions to everyone who attended. After looking around, all the available options were either outside of our price range, didn’t provide the set of features we wanted, or both. Several years ago, the decision would have been clear. Make a compromise and deal with a non-optimal solution. But that was then and this is now.
In a world where agentic coding tools exist, there is now a third way. Just build it. So I did. Instead of settling for one of the options available, I was able to create the thing I wanted from scratch, building it from the ground up to be exactly what I wanted. I started with a basic template, iterating bit by bit, prompt by prompt, to shape it into exactly what I wanted it to be. Want an admin dashboard? Ask Claude. Want to be able to export and import all the data into JSON files to back up or edit elsewhere? Ask Claude. Want to make it mobile responsive or add a feature for users to log in with magic links to track their personal schedules? Ask Claude.
There is, of course, expertise still embedded within this process of building. Although I didn’t write more than a few lines of code through the entire build process, I still was able to articulate what I wanted and to steer the AI coding tool (Claude Code, in this case) to design the site in the way that I wanted. This project was not outside the realm of something I could have done in a pre-genAI world. I know how to code and have been fascinated with technology for long enough to have a knack for figuring out these kinds of things and having a vision for what is possible and how it should look. But without AI this would have taken me at least ten times longer than it did, and probably closer to a hundred times longer. And that alone means it never would have happened.
Revisiting the AI Wisdom Matrix
And so, as we revisit the question “does it matter if AI was used to build it?”, the answer remains: “it depends.” As we grapple with the appropriate ways to use AI in the various spheres of our life and work, I’m reminded of the AI wisdom matrix framework I sketched out a few months ago. Although I’m not thrilled with the axis labels and it is probably due for a v2 makeover soon, I think that framework is a helpful guide. As you can probably smell, I used Claude to create those graphics too.
As we slog through the morass of “it depends,” we need to answer a more fundamental question: where does the primary value of the thing we are creating come from? Is the value embedded in the process by which it is produced? Or is the existence of the artifact itself where the actual value lies?
The reality is that the answer will often be some combination of the two, but I do think that in most situations, going through the exercise and trying to place the things we are doing on this matrix is enlightening. In the case of the conference website, my answer to the question is a resounding no, it doesn’t matter that AI was used to create it. The website was a means to an end, a tool to communicate a certain set of information to the people in attendance and a conduit to facilitate an in-person convening.
I don’t read Chayka (or indeed the folks he quotes in his article) as saying that using AI in design is always an abomination and that, for instance, AI should never be used to help build a pitch deck. But the Claude Design-generated decks that look like all the others are a signal of a certain misunderstanding of where that artifact should lie on the AI wisdom matrix. In many situations—and indeed in many of the situations that are most important to us as humans—a significant part of the value and signal of that product is embedded within the process that was used to create it. Outsourcing it to AI may give a polished product, but if that polished product is obviously the result of a lack of attention and care, then the veneer is thin and quite easy to see through once you’ve learned the tells.
In a world where we are increasingly using AI-powered tools in more of our work, we must continue to hone our discernment lenses so that we remain clear-eyed about the real meaning of our work. The new capabilities provided by AI enable all sorts of new creative output to come into the world, but it is a moral minefield.
If AI helps you create the thing that never before would have seen the light of day and that brings real value to someone, then by all means, use AI to create that thing. But if at the end of the day it’s gotta come from you, no amount of flashy, fancy web design or Claude-inspired color palettes will do anything more than taint the thing you created with the smell of AI.
Got a thought? Leave a comment below.
Reading Recommendations
PG from the archives on the Writes and Write-Nots.
Yes, it’s bad. The reason is something I mentioned earlier: writing is thinking. In fact there’s a kind of thinking that can only be done by writing. You can’t make this point better than Leslie Lamport did:
If you’re thinking without writing, you only think you’re thinking.
So a world divided into writes and write-nots is more dangerous than it sounds. It will be a world of thinks and think-nots. I know which half I want to be in, and I bet you do too.
This situation is not unprecedented. In preindustrial times most people’s jobs made them strong. Now if you want to be strong, you work out. So there are still strong people, but only those who choose to be.
Eric Hudson in Learning on Purpose writing about school as a gym, not a job, with three suggestions for how to make it happen.
Teach students the process and payoff of learning
Treat AI as an object of study
Use AI to create, not consume
The Book Nook
This week I caught up on a few episodes of LeTourneau University President Steve Mason’s podcast, Built for This. First up was a conversation with Scott Hartley about his book The Fuzzy and the Techie. In their conversation, Steve and Scott discuss why embracing both technical and liberal arts is so worthwhile.
Second up was my friend Michael James McGinnis, talking with Steve about his new book, The t-Shaped Engineer in the Age of AI, and what formation looks like in our current moment.
The Professor Is In
Today is July 1st, which means my sabbatical has officially come to a close. While classes don’t start for a few months yet, I’m looking forward to getting back into the classroom this fall.
Leisure Line
Couldn’t resist these bags I ran across at Costco a few weeks ago.
Still Life
A fun macro shot of a lizard I caught in the back yard earlier this week.
"slog through the morass of 'it depends'" is such an accurate description of practical ethics
Thanks, Josh, as always for helping me continue to think about the ways using these tools matter. Its seems to me that many people are working with the live questions that arise out of "it depends," most of which involve how. The matrix is a nice frame, so looking forward to seeing a v2 if it develops.
Your quote from Paul Graham at the end brought me up short as I have been spending much of the past year thinking about secondary orality and what it means to be living in a period of transition from a global culture dominated by writing technologies introduced with the scroll to one of digital flows of image and speech introduced with the infinite scroll. The frame of "the write and the write-nots" nicely illustrates the imperiousness of book-minded people, what Walter Ong called, literacy's tendency "to arrogate to itself supreme power by taking itself as normative for human expression and thought."
As a book-minded person, I'm looking for frames of acceptance treat literacy as one mode of thinking, and try to imagine others including those just coming into being as we use these new cultural technologies to think.