Silverback Gym Rebuild: Two Prototypes, $0/yr Hosting

Silverback Training Facility, a Brazilian Jiu-Jitsu, Muay Thai and kickboxing gym in East Hartford, CT, was running on Wix and the site had rotted. I ran a read-only backend audit against the live Wix API and public sitemap and found the business address was wrong in Wix's own business-info fields — it still listed the gym's old town, and that wrong address was feeding Google Business, Maps and every contact widget on the site. The online store was twelve unedited Wix demo products (a vase, a tote bag, a face serum) carrying real prices and Best Seller ribbons, live on the public site. The site was half-rebranded: the previous business name still sat in the title tags and meta descriptions of seven pages. Three separate pages — Classes, Instructors and Schedule — all shared one identical title tag, and the Contact page was published under the title Blank Page. Pages lived at nonsense slugs. There was no online booking, no pricing plans, no business hours configured, no h1 on the homepage, no social-share image on any of the nine subpages, and the contact form's data fields were labelled in French, left over from a template. None of it was cosmetic — it was actively confusing search engines and prospects.

I ran the engagement solo end to end: audit, content model, two competing prototypes, and a final delivered site. The audit produced a severity-ranked bug report covering both the Wix backend and the public-facing SEO. From that I built a single content model — one facts table (address, phone, programs, geo coordinates) plus per-page copy for eight page types — as the source of truth so neither prototype would invent claims. I then built Prototype A: targeted API fixes applied to a duplicate Wix site, never the live one, purging the demo store, correcting the address, adding Muay Thai to the business description, and setting business hours derived from the class schedule. In parallel I built Prototype B: a nine-page static site from scratch — home, about, classes, schedule, membership, instructors, contact, FAQ and a launcher page — built with real content from the live site and clearly flagged stand-ins for anything only the owner could supply.

Running two prototypes instead of picking a stack up front was the actual point. The client is not technical, and improve-the-Wix-site versus leave-Wix is a business decision about ongoing cost and control, not a coding decision — so I built both far enough to be judged side by side rather than asking the client to evaluate an abstraction. The Wix track proved its own ceiling: I mapped the full API surface and found business-info fields like address and hours are editor-managed and not writable over REST, and slugs, titles, meta tags and the rebrand have no API at all on a classic Editor site. Real fixes exist only through manual editor work. That finding alone was a decision-driver: it told the client what staying on Wix would actually cost in ongoing manual upkeep, versus a static rebuild.

The content model de-risked the from-scratch build by forcing every fact — address, phone, program list, geo coordinates — to be defined once and reused across the header, footer and structured data, so the build couldn't drift into invented copy. I directed the build through Claude Code: I owned requirements, the content model, design review, QA and the go/no-go decision on working against the Wix duplicate rather than the live site. The AI-assisted workflow handled scaffolding the nine pages against one shared design-token stylesheet, wiring per-page SEO (unique titles, meta descriptions, Open Graph tags, JSON-LD for SportsActivityLocation and FAQPage) and building the responsive grid, which I then verified by rendering every page at desktop and mobile widths and checking that grids collapsed, contrast held on dark sections, and every internal link resolved. The result is flat HTML and CSS with no build step and no framework — it opens in any browser and needs no server, which is what makes near-zero hosting possible.

The client got a real side-by-side comparison instead of a leap of faith: a working duplicate-site fix for the Wix option, and a fully built, verified nine-page static alternative, so the stay-or-leave decision was based on evidence rather than a sales pitch. The audit surfaced and named every structural defect — wrong address, demo store, duplicate titles, missing SEO fields — including the ones a Wix-only fix would leave partly unfixable through the API. The delivered static site is framework-free flat HTML that hosts for roughly $0 a year, against an ongoing Wix subscription.

Wix REST API, HTML, CSS, JSON-LD structured data, Claude Code, static hosting