Name a startup in an afternoon, not a month. Generate thirty candidates without judging them, filter on the four criteria that actually cost money later — sayable on a phone call, trademark-clear in your class, not fighting an established company in search, and not a mangled spelling of a better word — then check domain availability and buy the first survivor. The name matters far less than founders believe, and the weeks spent choosing it are the most reliably wasted weeks in early-stage company building.
Set a timer for thirty minutes and write thirty candidates. No filtering during generation — judging while generating produces six safe names and nothing surprising.
Pull from several wells so you don't get thirty variations of one idea:
Avoid the two patterns that date a company instantly: dropped vowels (a 2012 artefact of .com scarcity, now just a spelling tax) and generic prefix-suffix construction — Cloud-, -ify, -ly, Smart-, Get-. They aren't wrong, they're forgettable, which is worse.
One test worth running at the end of the thirty minutes: read the whole list to someone once, wait ten minutes, then ask them which ones they remember. The two or three that survive are the ones with a concrete image attached, and recall is the only property of a name that compounds — every other consideration on this list is a cost you avoid, not a benefit you gain.
Almost all naming advice is aesthetics. These four are the ones with an invoice attached.
1. Say it on a phone call. Read each candidate to someone and ask them to type it. If they need it spelled, you will spell it forever — on every sales call, every podcast, every conference introduction. That's a real, permanent tax.
2. Check the trademark, in your class. Search your national register and the USPTO for the goods and services class you'll operate in. Twenty minutes each. This prevents the most expensive rename there is: the one you do after you have customers, ranking and printed material.
3. Search the name plus your category. If page one is an established company in an adjacent space, you've volunteered to fight for your own name in search results forever, and you'll lose for years.
4. Check what it means elsewhere. If you'll ever sell outside your language, a two-minute search saves an embarrassing discovery later.
Thirty candidates usually reduce to five or six. That's the point of the filter.
Run the survivors through the checker below. It looks up nameserver records across the extensions founders actually use, plus the get-, try- and use- variants.
Read the results honestly:
The remaining gap is domains in a grace or redemption period, which can answer with nothing. So use the check to eliminate in seconds, then confirm the survivor at a registrar before you commit to anything printed.
On extensions: .com if it's clean and affordable, because it survives being read aloud badly. .io or .ai carry no penalty with a technical or AI-product audience — check the renewal price, not the first-year offer, because they're meaningfully more expensive to hold. .co is the best fallback when the .com is a squat you won't pay for. .app and .dev are HSTS-preloaded, which forces HTTPS.
Do not buy a mangled spelling to get the .com. A clean word on .io beats a dropped vowel on .com in every situation that matters.
Register the domain the same day you decide. A name you're still "thinking about" next week is a name that gets taken.
Then, in about twenty minutes:
Then stop. The name is done. Every additional hour spent on it is an hour not spent on the thing that decides whether the name ever matters.
kitty.build attaches the domain you bought to the site it generates, on your own Cloudflare account, and handles the DNS and certificate — the domain, the records and the repo stay yours.
Feed her an idea