I build what businessesneed to grow.Product. Growth. Web. AI.

24 years building across three countries. Growth, product, web, and 3+ years deep into AI. Whatever the business needs next.

Hello, what do you want to know about Rafael?

15+Tools & systems shipped
2Companies co-founded
$395K+Revenue from ads I manage
5xOrganic growth delivered
// Selected Projects
AI Agents2026

Narvi

Personal AI agent orchestrating WhatsApp, Telegram, Discord, Slack, and Gmail through a single interface, with per-group Docker container isolation and vault-injected credentials.

5channels orchestrated
Growth Tools2026

AI Risk Checker

Lead-generation widget live at cyberhaven.com/labs/ai-risk-checker: indexes 700+ AI applications with enterprise risk scoring, email-gated lookups, and conversion tracking baked in.

700+AI apps indexed
AI Agents2026

Content Machine

End-to-end AI content factory live at Cyberhaven: keyword research to published article through 13 specialized agents with quality gates. Used by the Content Manager.

13AI agents in pipeline
Growth Tools2026

Exit Intent System

Full-viewport lead capture with real PDF rendering, multi-trigger detection, 6sense personalization, and escalating suppression.

25gated assets in library
Growth Tools2024

KeyScope

YouTube keyword research engine with a proprietary 15-category coefficient system, 4-layer caching architecture, and 77% API cost savings.

15keyword categories with proprietary coefficients
Creative AI2025

AI Leaderboard

365-day animated bar chart race comparing 10 AI tools over time, with square-root scaling for compressed visual impact. Live at leaderboard.cyberhaven.com, built with vanilla JS and CSS, zero frameworks.

365days animated at 60 FPS
// Selected Writing
AI6 min read

The cheapest token is the one you never send

A free, open-source tool called Headroom shrinks the bulky context an AI reads before it is sent, so you spend fewer tokens for the same answer. What its own numbers say about where the savings are real, and a week of running it on my own setup.

AI5 min read

When the AI bill comes due

Uber burned through its entire 2026 AI coding budget in four months, and Microsoft quietly pulled most of its Claude Code licenses. Why the bill is becoming a forcing function, and what comes after exploration: not less AI, but a different mix per task, at a price that adds up.

AI5 min read

On SubQ, and why efficiency is the new frontier

Subquadratic announced their first model yesterday, and the launch tweet is closing on 10 million views. Why an architecture-level efficiency win matters more than another point on a benchmark, and what changes when frontier-level AI stops being priced out of reach for the regular developer.

AI5 min read

Why you’re hitting your limits in Claude

5-hour session windows, weekly caps, and context rot all eat the same budget. The context fee before you type a word, why I compact at 60% and not 95%, and how scoping MCPs per project cut my baseline.