Tactical Names

I spent years developing personal rules for evaluating domains: shorter is better, real words beat invented ones, and a name must be easy to spell. Manually applying these constraints to thousands of expiring names daily was impossible. Fatigue quickly sets in, causing you to rely on raw instinct and forget that a promising asset at $500 can be a devastating investment at $15,000.

I partnered with my buddy Claude to translate these implicit heuristics into a programmatic process. We quickly realized a simple numerical quality score was not enough. To make smart acquisitions, I needed an interface that could instantly estimate a realistic maximum bid, flag when to walk away, and explicitly detail whether a name holds more value as a resale asset or a business foundation.

I did not need a language model to evaluate every single domain. Most initial filtering happens faster and cheaper with conventional scoring. A scoring engine processes the daily file in fifteen seconds, sending only the top three hundred candidates to the AI. When the database lacks enough historical sales to support a valuation, the model purposefully withholds the number rather than guessing. I would rather show uncertainty than an un backed figure.

The interface is explicitly designed to help investors spot valuable names instantly. The first iteration packed the score, recommendation, bidding range, and time remaining into a single row, creating too much data to scan comfortably. To fix this, I stripped the primary list view down to the essentials: domain name, direct recommendation, and time left. Individual scores and deep textual explanations moved cleanly into a dedicated detail drawer.

After two months of iteration, Tactical Names worked but had become increasingly complicated. I experimented with two conversation-first versions featuring an active chat assistant, but neither felt right. After doing the math, I realized the conversational feature was incredibly expensive to run. The core value of the tool lies in the structured data already on display, so I cut the chat entirely and rebuilt the interface around a clean, data-first brief.

Rebuilding the product also allowed me to fix critical onboarding and micro-interaction friction. I replaced the standard marketing page with an instant trial where users paste a list without an account to preview ten resale and build opportunities. I also designed a separate Passed view to isolate automated engine rejections from personal human choices, and replaced the misleading Bid button with an I’m bidding toggle to log auction intent without implying an internal transaction occurred.