The handoff is getting smaller

Note

Note · 24 Oct 2026

The handoff is getting smaller

Parth Gupta · 2026

For most of my career, designing a product meant creating screens, prototyping interactions, documenting decisions, and eventually handing everything over to engineering. A lot of effort went into making sure the original intent survived that process.

Over the past year, this has changed quite a bit for me.

I've built an interactive document canvas, long-running AI workflows, a video generation pipeline, and a talking presenter. Much of the code was written by AI agents. My role was to figure out what to build, how it should work, and keep refining it until I was happy with the result.

I'm still a designer, and I don't consider myself an engineer. But being able to build has changed how I approach design.

The biggest difference is how quickly I can discover that my design is wrong.

Something that looks convincing in Figma can fall apart when it encounters real data, unexpected inputs, or interactions that weren't considered. Earlier, many of these problems would surface during development. Now I can discover them while I'm still figuring out the design.

But I've noticed something else, which I find even more interesting.

Building changes what you can imagine

While working on the video generation pipeline, I spent a lot of time experimenting with ways to animate a presenter. Lip-sync, facial movements, image generation, rendering. I explored approaches I barely understood when I started.

A year ago, I wouldn't have known that something like a mesh-warp rig could be used to animate a mouth. It certainly wouldn't have occurred to me while designing screens in Figma.

And that made me realise something.

As designers, we're often less constrained by how a system is supposed to be built. We approach a problem from the experience we want to create. Engineers tend to be more aware of technical limitations, existing architecture, and the cost of maintaining a solution. Both perspectives are valuable, but they lead to different possibilities.

In a traditional workflow, an engineer receives a design that's already been decided. They might never suggest an unusual approach because the team has already agreed on a simpler one.

When I'm designing and building at the same time, I can explore those alternatives before making that decision.

I can try something, see how it behaves, discover a technique I didn't know existed, and change the design because of it. Sometimes an implementation constraint leads to a better interaction. Sometimes a technical experiment opens up a completely new direction.

Building isn't just helping me execute ideas faster. It's giving me ideas I wouldn't have had otherwise.

I think that's an underrated part of this change. We're not just shortening the distance between design and engineering. We're creating a much tighter loop between imagination, experimentation, and craft.

What does this mean for designers?

There's a lot of conversation around the role of a design engineer. I find the term interesting, although I don't think learning to code is what defines it.

To me, it's about how far a designer can take an idea independently, while maintaining the quality of the experience.

I see this as four stages:

  1. Picture: You can communicate an idea through screens and visuals.

  2. Behaviour: You can make it interactive so people can experience it.

  3. Product: You can make it work with real data, edge cases, and actual constraints.

  4. System: You can define the rules and patterns that allow others, including AI agents, to build consistently without you.

Most of us have traditionally worked somewhere between the first two stages. That's largely how design teams have been structured.

AI makes the third stage much more accessible. The fourth is where I think things get particularly interesting.

When you're building with AI, you quickly realise that good taste alone isn't enough. You need to explain your decisions, establish constraints, anticipate failures, and define what good looks like. Otherwise, an agent can produce something polished that is completely wrong for the problem.

And when AI becomes part of the product itself, the nature of design changes again.

In some of the AI-native workflows I've been building, there's no longer a fixed interface for every situation. The system interprets an intent, decides what to do next, and generates an appropriate experience.

This means I'm spending more time designing things that don't look like interfaces at all. The intent, the underlying context and data model, the rules an agent should follow, its boundaries, and when a human needs to step in.

The interface becomes an outcome of those decisions rather than the starting point.

There's a lot more to unpack here, but I think we're gradually moving from designing individual experiences to designing the systems that produce them.

The value of good judgment

One thing I've noticed is that AI makes it surprisingly easy to fall in love with a bad idea.

A polished prototype used to take days or weeks. Now it can take minutes. The visual quality creates a sense of progress, even when the underlying idea hasn't been properly questioned.

This makes product thinking, taste, and the ability to evaluate your own work more important. We need to get better at deciding what deserves to be built, not just how to build it.

None of this makes engineering expertise less valuable. Building something secure, reliable, and scalable still requires considerable depth. But I think the boundary between design and engineering is moving. It's becoming less about who creates the design and who writes the code, and more about the depth of expertise each person brings.

I'm increasingly convinced that the designers who benefit most from AI won't necessarily be the ones who master the most tools. They'll be the ones who combine strong design judgment, curiosity, and enough technical understanding to explore ideas that previously seemed out of reach.

A useful question to ask yourself is: How far can you take an idea before you need someone else to carry it?

A year ago, my answer would have been quite different.

More importantly, I probably wouldn't have imagined some of the things I'm now able to try.