EntityMap — Chudi Nnorukam

I build web systems that AI models can read and cite. The skill underneath is attention: I find what an AI assistant sees, or misses, about a site, then close the gap so the right pages get recommended. 5+ products shipped solo, concept to production in days. chudi.dev is the public, measured proof, and the home of AI Visibility Readiness (AVR), the framework I built to measure why AI cites you.

Both implications are codified in the harness-evolution framework I publish at chudi.dev/framework. If you are building a multi-agent substrate, that framework is the recipe.

I run two AI agents on the same work. Bud, a Convex-backed automation agent, handles my LinkedIn scheduling, blog calendar, and overnight browser jobs. Claude Code handles my dev work, content drafting, and codex maintenance. Until this date they had never written to the same place; their decision-histories sat in separate stores, invisible to each other.

AEO (Answer Engine Optimization) is optimizing your content to be found and cited by AI search engines like Perplexity, Claude, ChatGPT, and Google's AI Overview. Not traditional Google Search. The core insight: Google ranks pages by popularity; AI engines cite pages by extractability. Those are different problems requiring different solutions.

citability.dev (DA under 10) achieved 15% citation rate, outperforming Ahrefs at 5%. The three strongest predictors were answer-first content, dateModified schema, and original data.

I used the AI Visibility Readiness (AVR) framework to run infrastructure audits on 7 websites. Each site was checked for 10 signals that AI crawlers use to discover and parse content: robots.txt, sitemap.xml, answer-first content, content freshness, structured data (JSON-LD), meta descriptions, canonical URLs, HTTPS, heading hierarchy, and social sharing readiness.

Perplexity cites roughly 6.6 sources per answer; ChatGPT cites only 2.6. Only 12% of URLs cited by AI engines appear in Google's top 10. These numbers mean your Google ranking strategy and your AI citation strategy are nearly independent problems that require separate solutions.

After enough failed experiments, I started noticing what did work. If I can't see everything in one place, it doesn't exist. Nested folders and pages are where my tasks go to die. The best day I ever had with a task manager was when I kept everything on a single Notion page. No navigation. No "where did I put that?" No clicking through three levels of hierarchy to find my actual work. One screen. Everything visible. That's it.

Based on the audit data and corroborating research, three factors had the strongest predictive power. Pages where the direct answer appears in the first 100 words get extracted more often. Pages with dateModified schema receive 1.8x more AI citations than pages without. Domain authority had zero correlation with AI citation rates.

Your AI citation rate is the percentage of topic queries where AI search engines name your site as a source. You can measure it manually in under an hour: pick 20 queries in your niche, run them in ChatGPT Search and Perplexity, count how many return your domain, divide by 20.

AI agents can read websites, but WebMCP can let them use website tools directly. In this guide, Chudi Nnorukam explains how (and why) he shipped WebMCP on two live sites despite near-zero adoption today.

A CME Micro Bitcoin Futures trading bot built within Bulenox prop-firm rules: Rithmic market data and order execution, a fade strategy validated in a quant factory of 606 backtest configurations, and position/loss-cap compliance.

Polyphemus is a systematic Polymarket trading bot: signal detection, execution with risk controls, and operational tooling, built and documented as a production-grade case study with Claude Code.

Chiron is the operating name for the full AI agent infrastructure built on Claude Code. It stores decisions, entities, and principles in a Convex-backed graph, routes tasks through a library of forged skills, and dispatches domain work to a fleet of standing expert agents — each grounded via the /ascend forging protocol before being deployed.

The Tradeify MES system runs a momentum-fade strategy on Micro E-mini S&P 500 futures through a Tradeify funded account. It is executed via Quantower on a Windows VPS, managed remotely, and observed through a dual-log model combining a strategy file log and Quantower Serilog. Contract scaling is evidence-gated, requiring n>=30 live confirmed trades before any size increase.

The Founder-OS Cockpit is a private progressive web application and FastAPI backend that lets the operator queue, dispatch, and monitor Claude Code agent jobs from any device. Jobs are queued through the PWA, picked up by a runner on the private network, executed as headless agent processes, and streamed back to the cockpit. The system is never exposed to the public internet.

The standing expert agent fleet is the capability layer of the Chiron harness. Each agent is forged through /ascend — a structured protocol that builds a glossary node, codex knowledge graph, grounding artifacts, and a SKILL.md before deployment. Agents cover SEO, MES futures trading, cold outreach, Convex, frontend motion, AI-visibility, and agentic control-plane work. A routing layer dispatches every task to the most grounded specialist.

Review Reply Copilot is a self-serve micro-SaaS that generates on-brand replies to customer reviews, built for near-zero maintenance and async value: email reports and batch reply generation, monetized through Stripe.

WebMCP (Feb 2026, Google + Microsoft) adds navigator.modelContext to browsers so AI agents can call your site's tools directly instead of parsing screenshots. Three files changed, one npm package added, 90 minutes, verified live on chudi.dev. No other SvelteKit implementation guide exists yet.

Dev docs are three files that persist task state outside the conversation. Before compaction, run /update-dev-docs. After compaction, say 'continue' and Claude picks up exactly where you left off. Plan.md holds the approved strategy, context.md holds key files and decisions, tasks.md holds the checklist.