An AI tool directory Kodeit designed, built and still runs. It is the same process we use for client work, on a product where we carry every consequence ourselves.
Most AI tool directories are long lists ranked by who paid or who launched last. We wanted the opposite: every tool judged on the same visible rubric, pages fast enough to feel instant, and content structured so search engines and AI assistants can quote it accurately.
It also had to run without a team. One person writes, scores and ships, so every recurring job either became part of the build or got cut.
Every tool gets a BigBang Score from 0 to 100: the sum of seven weighted sub-scores, computed at build time from the tool's own data file. The weights and the testing method are published on the site, so a reader can check why one tool outranks another.
Each tool, category and guide is a JSON file in git. Pages are pre-rendered at build, so a visitor never waits on a database query, and every content change is reviewable and reversible like any code change.
A command palette (Cmd+K) searches tools, categories and guides from an index generated at build time. It loads on first use, so it costs nothing on pages where nobody searches.
Every page type ships structured data: SoftwareApplication for tools, FAQPage, BreadcrumbList and ItemList where they apply, plus an llms.txt for AI crawlers. An internal audit page scores every public page for SEO and answer-engine readiness.
Newsletter and tool-submission forms are validated, honeypot-protected and delivered straight to Telegram. The form reports success only when the message actually arrives. Nothing is stored that does not need to be.
A post-build check fails the deploy if any page links to a comparison that no longer exists, so a content cleanup can never leave dead links live.
Lighthouse on the live homepage, October 2026: SEO 100, Best Practices 100, desktop performance 97.
Fixed scope, fixed price, and you own everything we ship.