I've just been reading Maggie Appleton's Planning with Agents, and it got me thinking about how much the process of making products is changing.

For a long time, the process felt pretty structured.

Research. Testing. Design. Build. Test again.

I would try to learn as much as possible before moving into the next stage, partly because moving into the next stage came with a cost. Building something took time. Changing direction took time. Getting something in front of people took time.

So I did the research before the design. I did the design before the build. I tested what I had built, then went back around if it needed changing.

That structure made sense.

The cost of exploring has changed

AI has started to break some of that down.

The cost of building a rough version is now incredibly low. The cost of getting information, exploring a possibility, or trying a different direction is much lower too.

I can have a thought, build something around it, see how it feels, realise it is wrong, and try another version in the time it used to take to get a first pass into a design file.

I find myself moving between research, design, building, and testing much more freely.

I might start by building a rough surface just to understand the problem. That immediately brings up questions about what needs focus and what needs highlighting, so I toss the prototype out for some rough testing and get something back.

Then I have another go with real data, which exposes restrictions I could not see in a clean prototype. All the way through, I'm adapting, finessing, and moving between design iteration, testing, and building.

All of that is something I can now do easily and quickly, without following a linear order.

The map is not the journey

Maggie writes about two kinds of navigator.

One leaves with a route already drawn. They have the map, the course, and a plan for how the journey should unfold. When the weather changes, they have to work out how to bring reality back in line with the plan.

The other leaves with an island in mind, but no fixed route. They read the wind, the waves, the birds, and the current as they go. The direction is clear, but the journey is worked out in contact with the world.

I like that distinction.

The second navigator is not less thoughtful. They are not just making it up. They have a destination, experience, and a way of reading what is in front of them. They just do not mistake the map for the journey.

That feels closer to how I want to work now.

I still need a direction. I still need to know what I am trying to make and why. But I am less interested in deciding every turn before I have started moving.

I want enough of a plan to begin, then enough room to let the work change it. Then I build, see what is actually there, and let the work and my own judgement carry it forward.

This does not mean skipping the hard parts

There is a version of this that is just chaos.

I can make a lot of things very quickly now. I can mistake that for progress. I can skip the thinking, generate ten versions, and still not understand the problem I am trying to solve.

The interesting part is that AI does let me skip parts of the process. Or at least skip the slow, formal versions of them.

I can learn something by building a rough version instead of spending days describing it. I can test an interaction while I am still designing it. I can explore a few routes before committing to one.

But it also brings the remaining parts closer together. I can make something, learn from it, and adapt while the idea is still alive.

I think I am still working out what that does to the role of design, research, and development. But the old sequence of research, then design, then build, then test feels less fixed than it used to.

And I don't think it is going back.