At Atom, a lot of what we’ve built over the years has been about reducing friction in how people find domain names.

Search got better.

Suggestions got smarter.

AI started understanding intent, not just keywords.

But there’s always been one step we haven’t been able to remove.

You still have to type something.

And in many cases, that’s the hardest part.

---

A small experiment

Over the past few months, we started exploring whether we could remove that step entirely.

Not improve it. Not optimize it.

Remove it.

Which led to a simple (and slightly uncomfortable)

What if the search box wasn’t needed at all?

That work led to an internal prototype we’re calling MindSearch.

---

How MindSearch works

Instead of relying on explicit input, MindSearch attempts to interpret intent signals and translate them into naming directions.

It’s a mix of pattern recognition, contextual inference, and language models trying to make sense of incomplete signals.

In simpler terms, it tries to figure out what you mean before you say it.

---

What we’re seeing so far

In early testing, the system is often able to identify:

• A general category (AI, SaaS, consumer, etc.)
• A preferred naming style (short, brandable, more descriptive)
• The level of clarity behind the idea

It’s not always accurate, but it’s directionally useful more often than we expected.

There are also edge cases where it picks up conflicting signals or leans too heavily into recent trends.

There are also moments where it becomes very confident… even when the idea itself isn’t.

We’re still working through that.

---

Where it helps

The biggest impact so far has been at the very beginning of the naming process.

When there isn’t a clear keyword yet, or when users don’t know where to start, MindSearch provides a first set of directions to react to.

From there, the process becomes more iterative.

---

Feedback and iteration

Each session includes a simple feedback loop.

Users can indicate whether the results were aligned or not, and the system adjusts accordingly in subsequent passes.

We’ve seen this improve alignment in multi-step sessions, especially when the initial idea is still forming.

Although in some cases, the system becomes more confident faster than the user does.

---

Availability

We’re starting a limited rollout of MindSearch on April 1.

To manage capacity, access will initially include a small number of free sessions.

---

Pricing

Additional sessions are available as part of a usage-based model.

Details will be surfaced within the product during the rollout.

---

Try it

MindSearch is now available as a limited experiment.

You can try it here:

atom.com/mind-search

---

A note

We don’t expect this to replace how people name things.

But if we can remove even one step from the process, that’s meaningful.

Even if that step turns out to be the keyboard.

---

MindSearch begins rolling out April 1.