It made me think about how I’ve been using AI lately as a kind of virtual advisory board. I’ll ask Claude Code to spin up a series of different personas (called “subagents” in AI lingo) to review what I’ve built or plan to build. I’ll ask for a designer to look at the user experience, a product manager to assess how all the elements fit together, a marketing guru to think about positioning. They’ll provide individual reports that Claude then distills to areas of overlap and unique opportunities from each perspective. I’m not fooling myself into believing that these are real experts; it’s all an amalgamation of what the LLM has gathered through its training data. But the feedback has been consistently helpful and often surprising. As a soloist, it’s a great way to bring in a proxy for outside perspectives, without having to set up catering.

I floated the idea that maybe AI should replace our boards, not our junior workers. It was an immodest proposal, one that got some laughter, some wistful sighs. But the threat of AI-related job displacement always points downward to people already at a disadvantage. Many are navigating financial insecurity, student loan debt1, and uncertainty about where their skills fit. Board members, meanwhile, tend to sit comfortably in their wealth and careers. And yet, if you told a director to cut costs, how many would think to fire their board?

Hearing those nonprofit leaders at the summit reminded me of a Museum Board Leadership research project I led in 2024, in partnership with the American Alliance of Museums (AAM). In it, we surveyed over 400 museum directors about their relationships with their boards, and found a lot of the same kinds of dissatisfaction. Now, there are plenty of good boards and committed board members out there. The best of them deliver on what I’ll call the “3 Gs”: guidance (i.e., providing expertise in areas that supplement and support leadership), governance (i.e., keeping the org to its legal, fiduciary, and ethical responsibilities), and gifts (i.e., fundraising).

But even effective boards take a lot of management. First, many meet a lot. Take a look:

A vast majority of museum boards, 88%, meet four or more times per year. Over a third (35%) meet seven or more times! And that’s just the formal board meetings, not the full slate of ongoing board interactions.

Close to half (43%) of directors spend between 11-20% of their time on board-related issues. In a 40-hour week (LOL), that’s roughly a full day. And a quarter of directors spend even more.

Ok, that’s all the math, English majors, I promise. One more graphic, though. We asked directors to grade their board’s performance, and boards to grade themselves. The average grades aren’t particularly impressive.

There’s a lot there to take in, but a few lines stand out: a C+ on providing guidance and support to the director/CEO, a C+ on legal and ethical oversight, and a C- on fundraising. That’s a lot of Cs on the 3 Gs (guidance, governance, and gifts).2

This doesn’t mean most boards are awful. It’s just that, for the time and effort spent on them, many directors aren’t seeing the ROI. There’s important work already being done on this front, and on the necessary diversification of boards, by groups like AAM and the Black Trustee Alliance for Art Museums. But the burden remains on directors and their staff to keep their boards informed, engaged, and pulling in the same direction.

Here’s where AI can genuinely be useful and make the experience less taxing. Board books alone, with their detailed agendas, departmental updates, and budgets formatted for easy scanning, take a ton of stressful hours to assemble. AI can do much of that in an afternoon. Same with post-meeting summaries, donor outreach plans, and other follow-ups on action items.

Ultimately, what I heard from those nonprofit leaders at the summit wasn’t really a desire for abolishing boards or replacing them with bots. It was resentment that too much of their energy flows upward to the people who should be supporting them, instead of inward to staff or outward to communities.

So don’t actually fire your board, or your junior employees, for that matter. Just “hire” some AI agents to do the grunt work and claim back time to do the work that actually matters to you.

1. Amodeis ex machina

Look, I’m not saying that the sibling co-founders of Anthropic and creators of Claude definitely read my little Substack. But here’s recently talking about how we’re in a “centaur phase” – a time when human-AI teams outperform either working alone – and how it won’t last long, which I wrote about here. And then here’s Daniela Amodei talking about how important studying the humanities will be as AI permeates more of our lives and work. “In a world where AI is very smart and capable of doing so many things,” she said, “the things that make us human will become much more important.” She also listed the “soft skills,” like empathy, communication, and critical thinking, that Anthropic looks for when hiring.

They’re really talking about two perspectives on the same phenomenon: one that sees what AI advancement will displace, particularly with jobs, and the other that sees what we can do to fill those voids as this happens.

The question is whether we’re ok with that tradeoff, especially as the market value of arts and humanities training remains unproven.

2. Which half is horse?

Sci-fi writer and activist (and fellow Cory) Cory Doctorow wrote about this tradeoff recently. He invokes “centaurs,” too, along with their opposite, “reverse centaurs.” If the former is a person augmented by machines, the latter is “a machine that is assisted by a human being, who is expected to work at the machine’s pace.” He cites an example from the Chicago Sun-Times, where a newsroom full of interns, journalists, and fact-checkers were replaced by one freelancer with access to AI. That freelancer oversaw the production of a majority of the 10 “Best of Summer Lists” published by the paper. In the “Summer Reading Guide,” as published, 10 of the 15 books listed did not exist at all. They were AI hallucinations, pushed out by chatbots that can work faster and looser than one single human if not kept in check.

Maybe we shouldn’t expect one person to do the work of 50, regardless of the tools at their disposal.

On tech, as much as it may seem like we’re at the whims of larger forces, Doctorow says “there’s nothing inevitable about it.” Meaning us centaurs still have some agency to decide which half is horse (so to speak).

3. When toil isn’t trouble

, a reporter at the tech blog Platformer, provides a case in point. She built a Claude-based agent she called “Claudella” and tested whether it could do her job.

In some ways, it could. Claudella was good at finding quotes Markianos may have missed. It also came close to mimicking her voice and wrote solid pieces after a few rounds of feedback. But it couldn’t land a joke, got erratic when given editorial feedback, and showed a favorable bias toward Anthropic when writing about AI topics (an understandable piousness towards one’s maker, maybe).

The experiment led Markianos to reflect on her writing process and how her thinking unfolds and clarifies through the drafting itself. “If I had Claude write my first drafts,” she writes, “even if I fact-checked them thoroughly, it would be a lot harder to tell whether the angle was my own view or the AI’s.” Her enthusiasm for the field also started to deflate. She enjoys the “drudge work” of deep research and iterative drafting, and seems less enthused by the interpersonal aspects of reporting that AI won’t replace anytime soon, such as talking to sources and networking with other reporters.

I find this latter reflection fascinating and challenging to my own assumptions and biases: what about people who don’t want to be more social, more connected? If some kind of prescribed way of “being more human” becomes the only solid answer to AI, for some, that could become its own kind of drudgery.

1. In this economy?

The “attention economy,” in which platforms and politicians commodify our focus, is giving way to something more invasive: the “attachment economy,” which cuts deeper, straight to our evolutionary need for connection. talked to Dr. Zak Stein about the concept as it relates to the growing trend of people turning to AI companions as pseudo friends or even therapists. Those companions, they write, are “able to hack the attachment system that shapes our identity and bonds us to others.” Dr. Stein says this is essentially giving AI companies “a backdoor into the human mind.”

Sci-fi futures tend to depict bio-hacking as about hardware, with implants turning people into cyborgs. But the attachment economy suggests a kind of software update has already begun, no surgery required.

2. Whose game are you playing?

The Center for Humane Technology calls AI systems “the ultimate game theory players,” always optimizing for strategic wins. Philosopher C. Thi Nguyen’s new book The Score explores how this kind of gamification hijacks our decision-making so that we humans optimize for likes, KPIs, and step counts, and the scoring of points becomes the point. He asks a poignant question: who designed the score you’re trying to maximize, and whose interests does it serve?

In the attachment economy, the games will only get more personal, harder to suss out who’s behind them, and harder to quit once we’re hooked.

3. Academic alarms

A new Brookings study warns that AI in education carries more risk than reward. A majority of students agree, with 65% saying AI "undermines cognitive development.” Educators are less concerned, possibly because most are avid AI users. The report draws a useful distinction: adults using AI tend to augment skills they've already developed (becoming centaurs, maybe), while students are still building those skills. It’s one thing to amplify your voice with powerful tools; it’s another thing entirely to let them speak for you.

4. Management 101

, reflecting on conversations he’s having with his students, says that what’s “hot in programming now” is “setting goals for various levels of delegation, breaking down work into tasks, clear direction giving, clear feedback, coordinating across agents... management 101!”

Lots of “soft skills” in there. Who knew they’d be so valuable? (I have some thoughts.)

5. Chat amongst yourselves

There's a new social media site just for AI agents called Moltbook. In that human-free zone, the bots created a religion and planned for a language only they could understand. Some people are freaking out, sure that we’re perilously close to a Skynet or Matrix-like reality. But others say it's more likely mimicry than conspiracy, and that the bots are just recreating the sci-fi insurgency stories they've been trained on. Still, if this does turn out to be the beginning of something sinister, it'll be fitting that it started on social media.

Enter Claude Code, a tool by Anthropic (currently only available only to paid subscribers) that lets you talk to your computer in plain English. You describe what you want and it writes the code that makes it happen. This can be anything from organizing your desktop to creating a complex, web-based app. I’ve been using it a lot, for things like building a new process for developing a theory of change (more on this soon!) and turning long PDFs full of paragraphs into a spreadsheet-formatted database. It hasn’t made me any more of a programmer than I was before, but it has unlocked something in a way that feels profound. It’s like a barrier has been removed, partly because I can ask dumb questions without feeling silly or judged. For instance, I needed to install some software (Python, as it turned out) but had never opened a terminal on my computer to run a command-line interface to do so. Claude walked me through it step by step, with a “Great question!” instead of a sigh or eye roll when I asked how to do it.

I’ve worked with some amazing, creative, and very patient programmers in the past, and many would likely have walked me through how to do some of these processes if I’d asked. But they were busy, and their job was writing complex code, not teaching me how to do a rudimentary version of it. This kind of distribution of responsibilities has been one of the main tenets of how work teams have functioned for at least the past few decades1. Now that some of this orthodoxy may be shifting toward “full stack builders,” Claude Code makes that kind of one-person-band work feel possible, at least for me.

Of course, there are tradeoffs. When I worked with human developers, there was often friction, and not just the frustrating kind, but the useful kind. They’d push back on ideas that weren’t fully baked and contribute their own from perspectives vastly different than mine. They’d help guard against “scope creep” when I thought about adding features. They’d have other projects, other deadlines, which forced us all to prioritize. And as great as Claude is at certain tasks, it’s not a very competent ping pong partner, nor one I can share inside jokes with that build over time and add texture to our bond (at least, not yet). Claude is faster and more readily available than any human, but, even with the occasional pushback or battles of will inherent in working with others, the lack of friction can be its own kind of problem.

I’m as reliant on Claude Code as I ever was on my programming teammates. Yes I can create things I’d never been able to before, but I can’t fix it on my own if it breaks. And if Claude goes down for maintenance or I run out of tokens, I’m stuck. Still, it’s not a bad time to not know how to code. It’s reassuring, at least, that even some of the most prominent names in programming, including those whose work has laid the foundation for current AI systems, feel “behind” as programmers.

These tools are unleashing my creativity and capabilities in new ways: if I can dream it, I can build it, or at least describe it clearly enough that it eventually gets built. Does it matter that I’m still leaving much of the labor to others? It strikes me that there’s an etymological link between the digits we rely on for manual labor and the “digital” work of computers: apparently since we humans have always used fingers and toes for counting2. So maybe the question is really about who or what we choose to count on, and whether we’re ok swapping the friction of human collaborations with the friction-less services of machines.

AI, there’s the rub.

Then I heard this podcast interview with Tomer Cohen, LinkedIn’s Chief Product Officer. In it, he talks about how, with the rise of AI tools, he’s no longer just looking for product or programming specialists for his team. Instead, he’s looking for what he calls “full stack builders”: people who can come up with an idea, create it, and communicate its importance and fit to their colleagues and clients.

“The idea,” he says, “is for that builder to develop experiences end to end, to combine skills and expertise across what was traditionally distinct domains to bring it all together.”

And the qualities he values most in his team are not Python fluency or creating killer product requirements docs (PRDs), but vision, empathy, creativity, communication, and judgment. I literally had to stop the podcast when I heard that and rewind it to make sure I’d heard it right. It was the first concrete evidence that this shift might be real. An actual executive at a major tech company, one whose whole business is helping people articulate their career skills, describing a move away from specialists and toward people with creative and communicative skills.

I wanted to send this to every career services department I’ve ever encountered with “SHOW THIS TO EVERYONE” scrawled in red. But I’ve come down a few pegs in excitement since then.

There’s still the challenge of demonstrating you can do this kind of work to actually get hired for it. When do you build that portfolio? Most people with demanding jobs and other life responsibilities don’t have spare time to experiment with AI tools and prove they can be full stack builders if given the chance. The space to develop and demonstrate these skills often comes from unemployment or underemployment, not from opportunity.

And what counts as proof? An art portfolio shows vision and judgment. A master’s thesis proves you can develop ideas end to end. But I suspect companies would want to see apps and other digital tools, which still (for the time being, at least) likely means that arts and humanities people won’t have the right credential to even get in the door.

It would also be worth knowing whether Cohen and others like him are actually hiring for these skills now, or just reshaping the roles of people they already have. And if asking people to do a little bit of everything is truly replacing specialists, what does that do to expectations around pay? Will generalists now get more compensation than specialists because they’re being asked to do more?

Of course, it’s not just a perceived skills gap that’s kept some people with arts or humanities backgrounds from working at these kinds of companies. There are plenty of reasons to be skeptical of Big Tech or startup culture, and it feels a bit icky to reduce people to their commodifiable skills. But people should at least know if these opportunities exist.

One person’s vision for the workplace doesn’t represent an industry-wide trend, but it’s worth paying attention to. For now, at least, there’s one concrete data point that suggests the rhetoric might be shifting toward reality.

Meanwhile, Meta recently approved a policy that lets their AI model talk to minors in sensualized ways. And the CEO of Roblox, in a recent interview, gave a “What, me worry?” response when asked about the lack of protections for kids on the platform. As , one of the interviewers, writes:

Given a chance to display empathy for the victims of crimes his platform enabled, or to convey regret about historical safety lapses, or even just to gesture at some sense of responsibility for the hundreds of millions of children who in various ways are depending on him, the CEO throws up his hands and asks: how long are you guys going to be going on about all this stuff?

The worst consequences we’ve seen so far for these companies are a few uncomfortable hours in front of a congressional committee or fines that are really just write-offs. And yet they’re basically running massive experiments on population-levels of subjects. They famously “move fast and break things,” even if those things are flesh and blood or whole swaths of the social fabric. They have ethics officers (apparently1), but no real ethical code.

What they have instead are lengthy user agreements that nobody reads before signing and “age restrictions” that anyone who knows how to backdate a calendar can get around. While social science research, to its collective credit, has gone to great lengths to protect individuals, these massive companies have crafted dense legal documents meant to protect themselves.

When it comes to potential profits or guarding against potential harms, these big companies have proven that they will always choose the former. Newton describes all the ways Roblox could have protected kids years ago. “But doing so,” he writes, “would require Roblox to focus on outcomes for children, at the likely expense of growth. And so here we are.”

Up until now, I’ve been focused on whether people with arts and humanities backgrounds have economic value in the age of AI. But there’s another argument here, one that might matter even more: we’re also needed. Not just for what we can offer the market, but for what we can offer as a check on it.

There’s a reason they’re called the humanities. Studies have shown that reading fiction increases empathy, that immersing yourself in stories changes how you relate to real people. The work of studying arts and culture isn’t abstract. It’s practice in seeing through someone else’s eyes, in understanding experiences beyond your own and caring about what happens to people. That kind of attention to actual human lives is what’s missing in big tech right now. We can’t leave the future of transformational technology to technologists and business people alone. We already know what the results will be if we do.

I learned this while working on the banner for this Substack. The different models I tried kept getting me most of the way to where I wanted to go, but ultimately I had to step in and do things “by hand” in a graphic design program (Figma). Turns out, this isn’t just my experience. Recent research out of Carnegie Mellon University shows that human-AI collaborations often produce better results than either working alone2. The current crop of AI agents are faster and cheaper than humans, but they can’t reliably produce human-level quality on their own. They approach problems with too much of a “programming mindset,” with an over reliance on the logical and systematic. When you’re a computer, in other words, everything looks like a program. As the authors put it, agents “take an overwhelmingly programmatic approach across all work domains, even for open-ended, visually dependent tasks like design.” Tasks for which we use our judgement, taste, and experience, like fine-tuning my banner image, for instance3.

The authors call this optimal collaborative state “co-intelligence.” But it’s not clear how long our necessity in the mix will last. There’s a historical parallel in the world of chess. In 1997, Garry Kasparov became the first Grandmaster to lose a match to a computer program, IBM’s Deep Blue. It was a turning point for AI in the public realm, and could have been a devastating endgame for Kasparov, but he used the loss to create a new kind of chess player: computer-human hybrids called “centaur teams.” By 2005, centaurs could reliably beat not just other human chess players, but chess programs running on their own, as well. But by the early 2010s, the competitive advantage of that version of co-intelligence had narrowed considerably due to advances in computational power.

I think we’re in a similar window now, where people and computers can augment each other’s abilities. Call it the Age of Centaurs. But it’s likely to be a short-lived era considering how quickly the tech is advancing and how desperately companies and countries alike want to be first to crack artificial general intelligence (or AGI), a state in which reasoning machines effectively don’t need us anymore. There are also clear economic pressures already reshaping the job market, with entry-level jobs harder for new grads to attain and companies like Amazon and Meta doing massive layoffs. As the CMU paper shows, current AI agents can complete tasks 88% faster and cost over 90% less than human workers. As AI quality continues to improve, the business case for co-intelligence will likely get squeezed out by the efficiency of the purely artificial kind.

In Pluribus, the infected gain all the knowledge of a hive mind but lose their individuality and free will. AI already has the knowledge part down, and will only get smarter. What it lacks are the messier, non-programmatic things like aesthetic judgement, gut feelings, and actually caring whether the output matters. Those are still ours. For now.

Of course, there are some (mostly) human-made movies that fail to pull me in already. Another one I saw recently, Tron: Ares, was one of them. It’s explicitly about AI and the dangers of tech advancing beyond our control, and has some interesting if somewhat muddled ideas, but the movie as a whole feels cobbled together like Frankenstein’s monster1. Despite the enormous effort CGI artists must have put into creating every scene, the final product didn’t feel too far off from what I imagine a fully AI-generated movie might be like. It has that kind of “auto complete” feel to it, leaning heavily on familiarity: action-movie tropes like egomaniacs setting off a chain of events that only the hero can stop, one-dimensional sidekicks, witty one-liners. The movie probably wouldn’t have even been made if not for the bet the studio placed on familiarity with the IP, even nostalgia for it, getting people into seats (which didn’t really work, but more on that in a minute).

While One Battle After Another felt like it was reaching for something idiosyncratic and unexpected, Tron: Ares felt like it was reaching for a formula.

Filmmakers relying on narrative formulas is nothing new, of course. From the hero’s journey to Save the Cat!, storytellers have long looked for successful or compelling patterns in other narratives and tried to emulate them. So maybe the dismissive “formulaic” isn’t too far off from “algorithmic.” In fact, algorithmic storytelling may just be a more efficient version of what studios have long been trying to do anyway.

Neither film has done particularly well at the box office compared to how expensive they were to make. One Battle After Another will ultimately lose less money than the Tron film, and will undoubtedly get love in award season, but its craft didn’t translate into wild commercial success. I may have loved it, but clearly my taste doesn’t move the market.

There’s a point in Tron: Ares where the protagonist, a tech innovator whose creations may end up destroying everything, says: “I should have majored in English Lit.” The joke being, I guess, that her life would have been so much quieter and less complicated if she’d chosen something safe and conventional instead of succumbing to her technological genius and changing the world. It was funny to hear that, partly because it’s right on topic for what this whole Substack is about. But also because the screenwriters who wrote that line likely studied liberal arts, yet they’re basically trashing their own areas of expertise as the opposite of ambitious and not what visionaries and world-shakers study2.

So, back to that post-viewing epiphany. I still think AI couldn’t make anything like what I experienced with “One Battle After Another.” And people like PTA and Leonardo DiCaprio, and other “above-the-line” talent, aren’t in any real danger of being replaced. What AI might replace relatively soon, though, or at least displace, are the hundreds of CGI artists, compositors, and technical workers who make most movies look the way they do. The “below-the-line” workers whose labor is already digital, already somewhat algorithmic. The very people who worked so hard on those impressive Tron effects. And unlike the writers, directors, and actors who have powerful unions that can shut down the whole industry, visual effects workers don’t have that same pull or protection.

We know the market wants efficiency. And the creative act is often messy, which is part of what makes it human. Even formulaic films take a collaborative effort that often blows through budgets and past deadlines. So maybe the question isn’t whether AI can pull off singular artistry like a PTA film. I still don’t think it can or will anytime soon. The question is who gets displaced in the process of making everything else. Because the market won’t ask whether we value human inefficiency enough to preserve it. The market will just choose efficiency.

I started with Claude, as it’s my typical go-to, and also the only platform I pay for monthly. My first prompt was pretty minimal, just the title of the Substack and a general sense of what I wanted to write about.

What came back was…bad.

It’s blocky and cold, like something from a ‘90s computer-training manual. And it came up with that lousy, inaccurate subtitle on its own. I didn’t spend a lot of time breaking down exactly why it felt off, though, and just trusted that my taste was right on this. Which in itself is worth noting: knowing something doesn’t work, even when I can’t fully articulate it, is an instinct I’ve learned to trust over the years. Claude also took the idea of a banner literally, and created something in html instead of an image format like png or jpg, which may have limited what it could do with design.

Then I tried the same prompt with ChatGPT.

This one was a little better and even has some of the elements that would show up in the final design, like the digital brain circuitry and the color scheme.

Still, it felt off and incohesive, like ingredients arranged on a counter instead of actually cooked into something. It was simplistic rather than simple, and even though I couldn’t picture what I wanted yet, I knew this wasn’t it.

Next, I tried Google’s Gemini, which I’d heard was good with imagery. The first result was way different than those from the other two models.

The photo-realistic library scene, the holographic Shakespeare, the floating digital brain, all kind of cool but there’s just too much going on here. Like Claude, it also added a subtitle that I never asked for and didn’t really like.

I continued to prompt in Gemini, going further than I did with the other two models, and it did something interesting. Instead of giving me another image, it gave me three written design concepts with visual descriptions and intentions of what the imagery was meant to evoke. Instead of just reacting to the next prompt and providing an answer, it felt like it was designing with me, offering options to react to and choose between.

Still, the process wasn’t a straight line to something I liked. Some of the iterations got closer to that feeling of “rightness,” but then the next would get further away. It was frustrating at times, I have to admit. Eventually, I realized I wanted something more like a logo, and an image that would look good as a standalone avatar or thumbnail.

Finally it came up with something I really liked: the split-brain illustration I’m using now, with circuitry on the left, in cool blues, and handwritten text on the right, in warmer hues. It still felt a little too busy at first, so I asked for a simplified version that took a few more prompts to get where I wanted it to go.

So I had the image, but still didn’t love the look of the text. I kept trying to prompt it to where I thought it needed to go, but that just got way too frustrating. It kept giving me weird font sizes and spaces between words, and insisting those issues had been fixed when they clearly hadn’t.

It was these little things, tweaks that should have been easy, that broke me and made me resort to doing it myself. I’d heard of the design and prototyping tool Figma, and wanted to try it out. It’s got some similarities to Illustrator and Photoshop, which I always found intimidating and not particularly intuitive. But I wanted to be able to manipulate the different design assets, and knew that using “layers” would help with that. It turns out the main difference between this exploration of Figma and my previous fumblings with design software was, yes, AI. Before, I’d have to search for answers to any small issue I was having or watch tutorials, which I could never replicate. Now, I just typed in the problem to my friend Claude and it (he?) would tell me exactly what to do. For instance, I couldn’t figure out why everything I exported was blurry and pixilated, and Claude immediately told me the setting that fixed the issue.

Here’s what I came up with (the Substack version has white text on a black background).

I’m pretty happy with it! For the text, I just scrolled through fonts to see what felt right, again trusting my instincts.

The whole manual process felt tentative but familiar. Unlike AI prompting, where I was trying to explain what I wanted and hoping the tool would understand, then getting frustrated when it didn’t, this was more direct. Tweak, step back, squint, repeat.

Ultimately, the AI gave me an image I really like and couldn’t have created on my own from scratch. The split brain with its circuit-board patterns and handwritten text, the balance between the extending nodes on one side and the ink splatters on the other, the composition that works at any size. I definitely don’t have the illustration skills to make that.

But getting from the AI-generated image to the finished banner required things that came from my background, not from knowing how to write prompts. Things like aesthetic judgment (i.e., taste), iteration, knowing when to switch tools, when to keep pushing, or even when to abandon something that wasn’t working.

Maybe it helped that I learned to make art through totally analog means, a process of making marks, adjusting, blending, and refining. This kind of felt like that. Moving across tools may not have been the most efficient process, but the messy back-and-forth was a creative process. And, just like any medium, sometimes the best way to gain confidence in the tools is just to dive in on a specific project.

So I’m curious: if you’ve used AI tools to make something, does this process look anything like yours?

At one point, the host, , stops him: “As one of the four people in technology that have a liberal arts degree, I have to call out…art literacy is really important right now.” I’ve heard similar sentiments elsewhere, that English majors, for instance, will prove more valuable in the coming job markets than programmers. That those best suited to take advantage of the new systems are people adept with language, critical thinking, and creative problem solving.

Vo goes on to say how key it is to be able to describe design, and that “language is now such a foundation for technology that if you’re not investing in your linguistic skills…you’re going to miss out.”

Oh, how I’ve invested! From studying art history, film, and writing, to working in arts and culture and the social sciences. I’ve spent years building up exactly these kinds of skills, ones that were rarely pursued with pure ROI in mind. I’ve also, at times, internalized the market’s judgment: that my pursuits were of less worth than more practical or STEM-centric disciplines, and that they therefore reflected something worth-less (though not exactly worthless) in me. So when I hear that these skills may become more sought after in the age of no-code, LLM-powered tools, I want to believe it.

I’m also skeptical. Maybe it’s because I really do have critical thinking–that highly valued skill–hardwired in me at this point. But I’ve also heard versions of this before. About generalists, for example, how they integrate more easily into workplaces than specialists, see across silos, and bring flexibility to otherwise rigid systems. Which sounds great and is often true, but try finding “generalist” in a job description, or comparing their salaries to those of specialists at the same level.

So my worry is that this is more rhetoric than reality. That we’ll be told we’re valuable while watching jobs and salaries go elsewhere. That the tools themselves will intimidate the very people who might thrive with them.

That’s a big part of what I want to explore in this space. If you’re an English major (or adjacent), does anything in this moment feel like vindication? Or do you think this is just another instance of people saying one thing but doing (or hiring/paying for) another? Let me know what you think in the comments below!