May 7, 2026 · 5 min read

The mom test

Every product decision Ezra makes runs through one filter: would my mom feel dumb here? It sounds glib, but it's the most honest design constraint we've found.

Most AI products are built for people who already feel comfortable with AI. The early adopters. The "vibe coders." The folks who already have a ChatGPT account, who type "Ignore previous instructions" as a joke, who know what an API key is.

That's a tiny audience. And worse, it's an audience that will switch to something better within a quarter, because they read TechCrunch and follow product accounts on Twitter and try every new thing that ships.

The audience that actually matters — the one that, if you can serve it well, you have a real business — is everyone else. The people who use iMessage all day, who manage their kid's school comms and their dental appointments and their work emails, who would benefit enormously from an AI that handles the small stuff but who would feel stupid trying to set one up.

What "feeling dumb" actually means

Software makes people feel dumb in specific, predictable ways:

None of that is unavoidable. All of it is a choice the product made.

How Ezra dodges each of these

No jargon, ever. Ezra never says "OAuth" or "API" or "agent" to a user. He says "let me see your email" or "tap once to connect Gmail." If a word isn't in your daily vocabulary, it doesn't appear in the product.

One thing at a time. Ezra never gives you five options. He suggests one next step. If you want something different, you can say so in plain English.

Confirm before consequential things. Before sending an email, scheduling a meeting, or anything someone else will see, Ezra asks. "Sound good?" Always.

Mistakes are Ezra's fault. If something goes wrong, Ezra says so in his own voice. "Sorry, I should have asked first." Not "User error: invalid input."

Errors stay in character. When something breaks, Ezra says "Something on my end — trying again." He doesn't expose the technical reality unless you ask.

Patient pacing. Onboarding takes as long as you need. If you're busy, Ezra says "no rush." If you're confused, he asks what would help.

Ezra's first message

The very first thing Ezra says to you, after you text him "Hi":

Hey! I'm Ezra. I help with email, calendar, reminders — that kind of stuff. What's your name?

Notice what's not in there. No "Welcome to your AI assistant." No "I can help you with a wide variety of tasks." No "Please tell me about your goals." No mention of the word AI at all.

Just: here's what I do, what's your name?

That's the bar. Every interaction Ezra has with you should clear it. If a moment in the product would make a person feel even slightly dumb — for not knowing a word, for picking the wrong button, for not understanding what's happening — we redesign that moment.

Why this is also a business strategy

The mom test isn't just a values exercise. It's the product's competitive advantage.

Every other AI product is fighting for the same 5% of users who are comfortable with AI. They're building features for a sophisticated audience that will move to whoever ships fastest. The other 95% of people are sitting there using iMessage all day, half-aware that AI exists, mostly intimidated by it.

If we can make Ezra feel as easy as texting your sister, we get a market that no one else is even seriously trying to serve. Not because it's hard to want, but because it's hard to commit to. Building for non-technical users means saying no to a lot of features that would make sophisticated users happy. We're saying no to those features on purpose.

Ezra is for everyone, including your mom. Especially your mom.