Robert García

I build and operate AI systems for the un-artificial. Not a developer who uses AI — an operator who directs it from the ground up: I decompose an ambition into a plan, assign the right grade of model to the right shape of work, set the checkpoints only a human may pass, verify the behavior instead of trusting reports, and measure what the work cost. AI writes the code. People own everything else.

I touched AI for the first time in March of 2026, and code for the first time soon after. In the short time since, that discipline has built and shipped:

• Two applications in daily real-world use — an AI triage desktop app I run my life on, and a job-search pipeline with a real client submitting résumés through it
• Two complete real-time strategy games — one packaged as a Windows executable with LAN lockstep play, balanced by an AI-vs-AI tournament fleet I governed
• A top-to-bottom cel-shaded graphical overhaul of Command & Conquer: Renegade — a 24-year-old game, open-sourced in 2025, restyled from a single one-shot /goal prompt
• A ~24,000-entry semantic library served over MCP — my own retrieval infrastructure, used across every AI surface I work in
• A published library of four agent skills — adversarially reviewed, A/B-tested, and privacy-scrubbed before sharing
• A governed multi-agent bug-eradication program — nine iterations, 12,449 API calls, hard human-only gates, closed on measured evidence with per-channel cost accounting

Everything on this site was built in collaboration with AI and says so plainly. The judgment, the doctrine, the verification, and the mistakes-turned-rules are mine. If your organization runs on models and agents — or intends to — that’s the skill set I bring.

Projects · Skills Library · The Oracle Blast — my answer to “choose which model?” · CV · The Letter — a commendation written by the AI I direct, provenance disclosed in its first line.

Contact: contact@robertgarcia.ai