Back when there was a Design Twitter, someone asked me: what should Design want? And I’ve been thinking about it ever since.

Individual designers want what most workers want: fair pay, and to be left alone so they can get some good work done. But of course we can’t get those things just by wishing for them really hard. Because: systems. And so my answer to the question of what should Design want — as, a practice, all of us together — is to create an environment where good design is possible.

I’ll be digging into some of it next Friday at Throughline conference, but there are two things that didn’t make it into the talk that I want to cover on the Picnic.

First: an environment in which good design is possible is an environment that nurtures a culture of maintenance. There is actually another talk at the conference on this very topic, by Ron Bronson — who wrote a piece on this theme last year. This is where those systems come into play: designers already do a lot of maintenance, but doing it within a culture that treats maintainers as janitors actively discourages those efforts.

Second: this environment should enable getting weird. We already talked a little bit about the inherent necessity of play and psychological safety for design to function last year, and the counterproductive erosion of our effectiveness by managers pushing predictable, linear delivery flows.

By now, readers of the Picnic might be confused, because I usually leave the “how to do it right” part for the end of the issue. But don’t worry: we didn’t just skip over the “how it’s being done wrong” section. Here it comes now.

The opposite of getting weird would be commodity-grade work: normalized, low value, large scale.

The opposite of a culture of maintenance would be a culture that valorizes velocity, of throwing things at the wall to see what sticks, and leaving what didn’t stick strewn across the ground for someone else to pick up.

If you guessed that I’m talking about vibe coding prototypes, you get a prize. What I specifically want to talk about is the system that vibe coding prototypes creates around itself. Even when the LLM does what it’s supposed to do (once in a while, the vibe coding slot machine really does pay out), how does it affect the great big social system we call work?

It turns that system into hell. Because while scrupulous professionals can get out of the addictive chat UI and deliberately evaluate the tools on the merits of getting important work done…that’s not how these tools are being adopted. Instead, they get picked up by executives hopped up on promises that they can ship their ideas right into production, bypassing all those tedious worker bees who used to be able to push back on bad ideas.

This has two important consequences. Firstly, while those managers fool themselves into thinking the AI is making everyone more productive, the reality is the opposite. Instead, the thoughtless introduction of LLMs into every process is a massive drag on environments that were already productive. Every expert’s job transforms into “workslop sifter.”

For any ordinary tool, people could say “no thank you” or — if someone else’s behavior was dragging down their ability to do work — raise it as an issue to a manager. But AI is The Future™, and you’re not allowed to refuse. Everyone is incentivized to slop as hard as possible by the very management whose job it was supposed to be to stop this.

And as productivity falls, the second consequence rears its ugly head. Managers feel the pressure to release something. The phantom velocity of the slopotype becomes the only thing capable of catching up and making the output metrics look good again. Teams give up on problem framing or divergent thinking. Assumptions — rather than being carefully interrogated (which is hardly done well even in functional environments) — become established as laws of physics.

Unfortunately, the trap that teams are falling into has irresistible bait. Prototypes are flashy and instantly appealing. Vibe coding lets you quickly do a thing that used to take a long time. The combination feels like a dream come true, even as the focus and rigor of your team unravels.

But none of this is inevitable.

One thing we can do more of is figuring out what the stakeholders are really asking and changing the conversation to focus on that. Forget the prototype for a second; what question are you trying to answer with this, or what decision are you trying to make?

LLMs can also be a “contrast dye for where resources are lacking” — they show up where people are being asked to do too much, without appropriate support. Without the LLM, the problem would express itself some other way. Is it possible to provide your colleagues with the necessary resources to do critical work well, and simply wind down the work that was never important to begin with?

These are just some of the ways we can start repairing the system that AI has wrecked, and getting back to a place where it’s actually possible for the field to move forward.

— Pavel at the Product Picnic