Dive Club
Posts
🤿 apple → airbnb → AI

🤿 apple → airbnb → AI

how the value prop of design is shifting

Ridd
April 19, 2025

Tuhin Kumar has designed at some of the most prolific companies in the world (early Facebook, Airbnb and Apple).

But everything’s changing in his new role…

Because now he’s the Head of Design at Luma AI which is one of the companies leading the charge on generative AI.

So this episode goes deep on how the value prop of design is shifting.

Some highlights:

The #1 trait of craft-oriented companies
Why you can’t think of AI products as just a tool
How to invest in the right skills as a designer today
What created Tuhin’s career tipping point at Airbnb
What Tuhin is pulling from Apple and Airbnb’s culture
How working with AI as a material is changing everything
The biggest lesson Tuhin learned while writing system prompts
a lot more

Listen on YouTube, Spotify, Apple, or wherever you get your podcasts 👇

🤝 WITH PAPER

I’m all-in on Paper as the next great design tool.

And this is coming from someone who’s taught Figma to more people than just about anyone 😅

It’s time for a tool that puts creativity at the heart of everything—not systems and process…

The way we design and create software is changing faster than ever and the tools we’re using today were made for yesterday’s tasks

Instead Paper is putting designers first.

So if you want to be one of the first to use the next generation of design tooling. Click the link to get on the list. Early access launches in May 👇

🎓 KEY TAKEAWAYS

My notes from chatting with Tuhin Kumar

1 — Art of writing system prompts

Remember in Tuesday’s episode how I said my project for this week was to work on system prompts for Inflight?

Well turns out it’s kinda hard 😅

You end up wanting to add a million conditional (if this then that) statements. But after going down that path a bit I remembered something Tuhin shared in our interview:

“you're putting a ceiling on the intelligence yourself because the way you wrote those logic pathways and those rules are the limits no matter how intelligent the model gets.”

Tuhin Kumar

So I reverted back to my simpler ideas in an attempt to avoid a similar rewrite down the road. Perhaps a bit of short-run wonkiness is the price you pay for being able to take full advantage of the new models. But I’m learning these lessons on the fly :)

2 — How to think about what skills matter today

Sometimes it feels like every episode tacks on another skill that designers all of a sudden are expected to bring to the table…

Here’s how Tuhin thinks about it:

Visual design is still the baseline. Having good taste and being able to execute on it is the foundational skill of design.
It’s impossible to be great everything (he mentioned how all of the best designers in the world were at FB in ~2013 but each person still had their strengths/weaknesses)
Everything else is an “amplifier” — you have to pick the things that are most interesting to you
The value of understanding AI as a material is similar to when a subset of designers got really into prototyping after swipe gestures became possible ~10 years ago. It wasn’t necessary per se, but the designers who could prototype went further.

3 — The design process is more spread out now

Tuhin talks about how he remembers projects at Airbnb or Apple where he would spend weeks thinking about the right design and iterating…

But now that “tidy linear path” as Julie Zhou refers to it is a thing of the past.

Design is still responsible for coming up with ideas, but so much of the new process of design revolves around figuring out the capabilities of these models.

So I asked Tuhin what it’s like collaborating with an in-house research team:

“this is not the normal engineering that we’ve seen for the last 15 years… I think this is the first time design has met a function that is even more timeline averse”

Tuhin Kumar

Instead of designing mocks for engineering to implement, Tuhin’s design process often looks like creating a prototype that “shapes what research should even research”. And what you ultimately design and polish is directly tied to the output of that discovery process.

I don’t have an in-house research team but I’m already feeling that same dynamic with Inflight.

For example, after you send some WIP for feedback I thought it would be nice to see results in a tidy Insights page that might look something like this:

But this page is almost 100% non-deterministic. So at the end of the day I only have a low-fidelity idea of what is possible 😆

If I’ve learned anything from these interviews it’s that when you’re building AI products you have to “prototype and prune”. So we shipped a code version that looked like this 🙈

But here’s the thing… tinkering with the output of these models has helped me discover possibilities I never would have on my own!

Which is something that Tuhin shared as well. Even these concept pills (one of my favorite features in Dream Machine) were an unexpected product of Tuhin simply playing with the models 👀