July 1, 2025, 9 a.m.

Initial notes on post-LLM design

Draft's Letters

Our prices on new consultative work will change at 5p CDT today. Now’s your last chance to buy a teardown, roadmap, or call with us at our current rate. If you’re an active client, your fees will not change for the duration of your work with us.


A couple of weeks ago, I surveyed everyone about what we should write about next, and “post-LLM design” won by a landslide. I find this to be fascinating, because we’ve positioned ourselves rather firmly as skeptical of LLM, at least with respect to some of the messaging that exists around it.

For the first half of my career, tech was changing the world, and there was a lot of optimism & promise around it. We didn’t really question much. The iPhone came out, then we all made a bunch of apps, then we ruined taxis & hotels. The internet was a place where we could all connect with one another and understand the world’s information. Twitter had infighting, but we thought it was funny.

2016 began to change a lot for tech. Two major democracies backslid. The rumblings around radicalization & abuse started to become more public. Over the past decade, we’ve mostly watched the promise of tech’s future slowly curdle. Does anyone think tech’s best days are ahead of it now? Is anyone looking to a new product to “save” us? Do customers trust us or care about us? High-quality people are too busy hiding in the dark forest to answer these questions.

It’s in this climate that people are in fact trying to find the next big thing. First we tried straight-up grifting, and then we mostly stopped when it got exposed as straight-up grifting, but not before the grifters got a lot of money and donated it to fascists. Now we have fascism & quieter grifters. Cool.

And now LLM is here, which is strictly more useful than grifting, but is certainly overhyped. When we see people speak of a new technology in near-religious tones, yet few people seem interested in using it, we start to ask why, and who stands to benefit from it. Meanwhile, no ethical person has enthusiastically clicked a sparkle emoji in the past three years of interface design, and more people are worried about LLM taking their jobs than what it can do for them.

We don’t think LLM is going to kill design. We also don’t think LLM is going to help design as much as most people think it will. LLM looks like it’s more poised to help developers be better at their jobs, and less likely to have a similar impact for design. (Almost all “best practices” were solved for a decade ago.)

That all said, we remain deeply concerned that those who typically buy design will become skeptical of doing so because of the narrative surrounding LLM, and not because of any attributes inherent to design or its proven impact. LLM is fundamentally a labor issue. If you don’t view it as a labor issue, we’re unlikely to convince you otherwise. If you have been religiously converted to the other side, we’re unlikely to pull you back to earth.

That might make us more qualified to write about all of this, though? Lord knows you aren’t seeking us out to teach you how to write better prompts. We care far more about:

  • What the technology really is
  • What the technology can do and (probably) will do
  • What knock-on effects the technology will have, both widely and for our own practices
  • What buyers of design will come to expect from us as a result of the technology
  • Any shifts to the relationship between consultative, strategic work and production work

The results of these questions are probably more evergreen than anything I could possibly write about prompt engineering or increased productivity. It is also probably more useful. Right now, designers need to know more about how to be than what to do.

You’ll figure out what to do. We all do. I’ve been doing this for 20 years, and I’ve more-or-less gone through five separate jobs in the course of my career. My role is utterly unrecognizable from when I graduated college in 2004. I don’t know how long you’ve been in this, but I expect similar transformations to take place for you.

How to be is much harder. It involves deeper questions about what you value, why you show up to work, how we can reason more deeply, and how we can all be more kind & compassionate to one another in an increasingly burning world.

If you’re interested in post-LLM design as a way of answering how to be, you’re in the right place. Otherwise, I may as well teach you how to run a customer interview, which got second place in the poll? If any of this resonates with you, hit reply. I feel like writing about this is both a tremendous distraction from my actual design practice and potentially the thing that people care most about right now, so even a single sentence or thumbs-up emoji helps give me the confidence to move forward with this work. Let me know.

You just read issue #212 of Draft's Letters. You can also browse the full archives of this newsletter.

Draft