The best tool is the one you don't have to use

I have three email inboxes. I check one of them.

Between a design studio and two kids, every day brings a school request, a client thread, and a calendar that never quite matches reality. The other two inboxes sit there until something goes wrong.

A few days ago I started letting a personal agent handle my email, my calendar, and some bookings. It reads all three inboxes, searches them faster than I can remember what I’m looking for, and puts things on my calendar without me typing a thing.

Then it did something that made me stop.

All I gave it was the name of an event, and it told me the ticket price. Before I could book, it was back with something I hadn’t asked for: another website was selling the same tickets for a little less. Go with that one, I said. It bought them, paid, and sent the confirmation to my email.

I never opened either website.

The DMV’s website is one I did open, on my own, and opening it was the problem. Renewing my driver’s license there, I got completely lost trying to figure out what to click. With the agent, I answered a few questions in the chat, approved, and it was done.

For me, if the agent can solve a painful problem that has been sitting there for so long, I am totally fine handing my card number and login to him. Besides, it always assures me that my info is safe. I gave it a try anyway.

Both times there was a tool between me and what I wanted: a ticket site, a DMV form. I didn’t have to or want to touch either one.

This thought has been hanging with me for days: the best tool is the one you don’t have to use.

It argues with something I wrote in an earlier letter, that the interface isn’t dying. I still half believe there’s value in interfaces, and I’ll come back to why.

What future apps are competing against

Since Meta’s muse came out, I have been testing it against Instinct (the one bought my tickets from the above examples).

On paper, Muse should win. It’s a full app with tabs, settings, an invite page, to-do lists, and a personalized feed. You can set health goals, financial goals, relationship goals. Everything is there.

But, that’s exactly why I stopped opening Muse less. Every tab is another place for my attention to leak, and more cognitive load thus less desire to use. I’d open the app to ask one thing and find five other things asking for me.

Instinct is just a text thread. That’s it. It lives in iMessage, the same place the people in my life text me, and that changes what a ping from it means. A notification from an app feels like one more product reminding me it exists. A message in iMessage feels like someone just texted me.

My first guess was that Instinct wins because I can only do one thing at a time there. That’s not really true. I multitask in that chat constantly. Oh, my son’s school needs this, can you handle it. Oh, and this other thing too. My brain runs five threads at once, like any parent running a business.

The difference is where the options come from. In Muse, the product writes the menu. In a text thread there is no menu. Every request starts in my own head, and nothing on the screen competes with it.

Some menus I would miss: ordering a drink, browsing a dress, and picking my son some toys. When choosing the focal point, the menu stays. A DMV form or three inboxes are hassles. Nobody enjoys choosing their way through a chore, and for a chore the best interface is the one that disappears.

Instinct barely has an interface, but it’s the one I keep going back to. I can tell you it’s not just the 0 UI that’s captivating, there’s more.

Why memory and reliability builds trust for Ai products

If a thin interface were enough, I would still be paying for Stanley.

Stanley is the AI I told you about in an earlier letter, the one that manages your presence on X. I’d read that it hit $3M in a few weeks, tried it out of curiosity, and signed up for the paid plan the next day. It worked like a text message too. I could say post this next week, and it just figured it out.

It was also genuinely good at the hard part. My old posts taught it how I talk: my tone, how I structure a message, what a hook and a headline should look like for me. Honestly, a lot of human assistants never got that.

The small things are what went wrong.

I told it I don’t work on weekends, so please don’t send me anything. Every weekend it pinged me anyway, the same routine as any weekday. I kept telling it. Nothing changed, and I had no way to stop it.

Another example that it lost my trust: one night I found out it could connect to my Dropbox. It said yes: it could see the folder and would work through the documents whenever I asked. I went to bed happy. By morning the task hadn’t run. When I pushed, the answer changed. It couldn’t read a single file in that folder.

Stanley never lost me over something it couldn’t do. It lost me over a rule it forgot and a yes it didn’t keep. Oh yes, I tried this with Instinct, and I was happy to say it holds its promises and automation well!

With Stanley at $80 a month and still pinging me on weekends, I gave Instinct access to my social accounts and my past writing and asked it to do the same job. It figured out my tone and writing style quickly, and it wrote that style down so it could keep using it. Plus, It’s free so far, maybe they are still testing with early users and not at a phase for monetization just yet.

Look, Stanley was built on solving social media writing and posting problems, yet it cannot remember a Saturday rule. Instinct isn’t built for social media at all, but it figured out more and followed more of what I have asked for.

We used to say niche down is the best strategy for a business. With Ai, I really doubt this statement is true now.

A general agent matched the skill on day one. We ask basically the same thing of a human assistant: follow through to the end, without needing to be chased or reminded. I have more confidence that Ai is more and more able to replace human assistant.

What’s left to design when there’s nothing on the screen

From both examples, there isn’t much screen left to design. One lives in iMessage and the other works like a text message. The UI is collapsing into places people already live, and users don’t have to carve time out of their day to learn another tool. That part of my job is shrinking, and I’m okay saying it out loud.

The UX moved. It lives when the agent speaks up and what it remembers, the two things Stanley got wrong. How it sounds is part of it, and so is what it asks before it spends my money, the way Instinct came back to me before it bought those tickets. Zero UI isn’t zero design. It’s design you can’t screenshot. I am really curious how Instinct product designer contributed to its launch.

In previous UX process, many designer starts with a persona: who they are, what they want, what frustrates them. But just from my own example: client email in the morning, a school request at lunch, something about a relationship at night. That is three personas in one day, and all of them are me. How about millions other users?

Nobody could have written that user down in advance. A personal agent becomes a product for exactly one person, and you meet that person only after you ship, once they’ve used it their own way and you can see the patterns.

How are we suppose to design now when persona are not defined since the beginning?

What I’d take from this as a designer

Count what your product leaves out. Every tab, feed, and setting is a place for attention to leak. If an agent can do the work, the screen should get smaller, not bigger.

Run the Saturday Test before you ship. Pick the smallest rule a user could give your product and check that it still holds a week later. Trust is won or lost there.

Design for the day, not the pain point. Map the common clock your users share, then decide where your product should speak up and where it should stay quiet.

For designers, the work is moving from what people see to what they can count on. Instinct is only a few days into its relationship with me. It hasn’t faced a Saturday yet. I’ll let you know how it does!

Studio Salt

I run Studio Salt, a fractional design partner that serves early stage startups.

Share my newsletter

Loving my content so far? I’d appreciate if you can share my newsletter to a friend 🙂