How I Built This Website: Zero-Framework Speed, Operations Rigor, and Pair-Programming with AI

By John Goliash · October 16, 2026 · 6 min read · Web Architecture & AI

Why an operations analyst hand-crafted a zero-framework, $0-per-month web architecture on Cloudflare Pages—and how pairing deep operational domain experience with modern AI created something faster, leaner, and more resilient.

1. An Operations Mindset Applied to the Web

In manufacturing plants and distribution warehouses, unnecessary waste is the enemy. When an ERP system like Sage or JobBOSS says there should be 50 units in bin B-04, but a physical cycle count finds only 48, you don’t simply adjust the number and walk away. You trace the discrepancy to its origin: Did a picker pull the wrong item? Was an outbound shipment never posted? Was an inbound receipt recorded in the wrong aisle?

In operations, we value precision, clean baselines, and eliminating waste. Yet when you look at the modern web, almost everything runs in the opposite direction. Simple portfolio sites regularly load 5 megabytes of JavaScript, thirty nested npm packages, heavy framework runtimes, and bloated tracking libraries—just to display plain text and a contact form.

When building john.goliash.com, I decided to treat web architecture with the exact same rigor I bring to inventory and ERP reporting: keep the baseline clean, eliminate waste, and build everything with the end user in mind.

2. The $0-a-Month Architecture: Zero Frameworks, Total Speed

This site runs without React, Next.js, WordPress, or Tailwind. It is written in pure, hand-crafted HTML5, modern CSS3 (using OKLCH color spaces), and vanilla ES6 JavaScript.

Here is what this architectural choice achieves:

Cost: $0.00 a month. Hosted globally on Cloudflare Pages' free tier, distributing static assets to edge data centers worldwide.
Lighthouse Score: 100/100. Instant page loads under 200 milliseconds, with zero layout shift and zero render-blocking framework delays.
Zero Maintenance Churn: No weekly security patches for vulnerable third-party npm packages, no breaking framework major-version updates, and no server database crashes.
Strict Privacy: Zero invasive third-party ad networks. Google Analytics and Google Tag Manager operate strictly behind an explicit user opt-out switch and respect Global Privacy Control headers.

3. Interactive Data Engineering in the Browser

A resume can say you understand KPIs, but interactive software proves it. The site features two substantial browser-based data applications:

The KPI Lab: An interactive supply chain analytics application built right into the home page. It models 14 operational metrics—from Inventory Accuracy and Fill Rate to Overall Equipment Effectiveness (OEE) and Scrap Rates. Visitors can adjust sliders, inspect live formulas, read teaching notes on why each KPI matters to plant profitability, and download client-side CSV files. It also packages into an offline Electron desktop app for Windows, Mac, and Linux.

Animal Crossing Villager Data Reports: An end-to-end data auditing project joining 391 villager records across three public Kaggle datasets. The analysis uncovers hidden data discrepancies—such as 260 song disagreements between public files and missing join keys—visualized through custom client-side charts without uploading data to any external server.

4. Craftsmanship in the Technical Details

Great operations work happens in the details. The site incorporates several deliberate design and engineering solutions:

Continuous Green-Bar Paper Résumé: A visual tribute to vintage IBM continuous-form computer paper with alternating greenbar rows and simulated tractor-feed margin holes that advance dynamically as the visitor scrolls down the page.
OKLCH Color Science: Using modern CSS color-mix(in oklch ...) color calculations to avoid muddy purples or near-black dead zones when blending blue and amber tokens during dark and light theme switching.
Single-Page Résumé Print Engine: Standard web pages often break across unwanted pages when printed. I built a custom print engine using clean Letter dimensions (7.95 in × 10.45 in) that prints a single-page physical copy with uniform thin bezels across desktop browsers and mobile devices—clearing Apple AirPrint's mobile Safari bottom URL footer without spilling onto a blank second page.
Structured Semantic Knowledge Graph: Schema.org JSON-LD structured data with explicit Person, WebSite, and FAQPage graphs, paired with an llms.txt site guide and RSS feed.xml for next-generation AI search engines.

5. Pair-Programming with Modern AI

I didn't use AI as a magic button to dump generic code. Instead, I used advanced models (Claude Opus 5.5 and Google DeepMind Antigravity) as active pair-programming partners.

The division of labor was clear:

Domain Expertise & Truth: I provided the real-world operational logic—the exact reasons inventory numbers drift in manufacturing plants, the specific reporting needs of purchasing teams, and the architectural constraints of zero-framework static deployment.
Implementation Velocity: The AI assisted in translating those requirements into clean vector SVG geometries, automated deployment staging scripts in PowerShell, and rigorous unit checks.
Human Verification: Every single line of code, formula, and visual breakpoint was audited, tested across real devices (iPhone, Mac, Windows, Chromium), and vetted against automated deployment gates.

6. Looking Ahead

Whether optimizing a warehouse floor plan, configuring an automated reporting schedule in KnowledgeSync, or building a high-speed website, the goal is always the same: measure twice, cut once, and make sure the numbers are right before anyone acts on them.

I am currently seeking my next career opportunity in data analysis, operations systems, and reporting leadership across Northeast Pennsylvania and remote roles. If your team needs someone who bridges raw database numbers with practical plant floor reality, let's talk.