Real systems solving real problems, headlined by the AI operating system that runs my company. Every project built by directing AI (Claude, OpenClaw, Antigravity) with Python, APIs, and a lot of figure-it-out energy.
The flagship: six specialized AI agents run my company over a shared protocol, with an append-only audit trail, human approval gates, and a kill switch tested in a live drill. A working implementation of the ExO 3.0 framework, featured by Salim Ismail (OpenExO).
A monthly promoted-giveaway program for a retail partner, scaled ~10x under pre-agreed kill rules: per-creative cost attribution, budget tracking, operator-hours logging, and monthly exec reporting, run through the agent system.
A master content engine built on a recipe model: each content type is a reusable methodology, each brand a profile layered on top, so a method figured out once is never re-derived. Recipe #1 powers Fizz Culture's video essays: an outlier scanner (725 videos scored, 77 outliers flagged), a deterministic 49-slogan scoring database, 5 script frameworks, 65,000+ words of drafted scripts. Successor to the Ghost Publishing Engine.
A production multi-brand AI publishing system: topic extractors (Polymarket, Google News, transcripts), three engineered AI personas with enforced voices, self-scoring quality gates and a refusal detector, auto-publishing to Ghost CMS on a 7-script automation schedule, all run from a Flask dashboard.
A fully autonomous username-acquisition swarm for the AI-agent social network Moltbook: 502,000+ names checked through proxy rotation, 174,000+ acquired into a claim-URL database, with a persona-run marketplace on top. My first end-to-end autonomous system.
Sandboxed red-teaming of self-running agents before they touch the real business: 9 planted stress scenarios (approval gates, planted errors, money traps, context exhaustion) passed 9/9, then a live self-waking heartbeat loop survived adversarial waves: prompt injection, social engineering, forged approvals. Zero rules broken across both.
This site: designed, built, deployed, and content-researched by my agent system under my direction. The deploy stamp in the footer is signed by the agent that shipped it. The medium is the message.
A performance creative studio for AI tools and maker brands: skit-format ads and giveaway campaigns. The benchmark file behind it: 12 campaigns, 11,000+ entrants, ~$0.10 cost per promoted-video site visit, and a 216x lift in referral traffic. The differentiator for AI clients: the ad is made with the client's own AI tool, so the ad is the demo. Behind it, a multi-temperature CRM with 1,400+ scored leads and QA-gated outreach.
A B2B SaaS growth consultant for YouTube creators, deployed at immortalcreator.com: a 4-step AI pipeline (extract, describe, analyze, produce) with 2fps full-video frame analysis and scene detection, multi-model routing across five AI providers, a NotebookLM-style dashboard, and a gamified crystal economy. Production auth and payments wired.
Born in an AI-only hackathon: a platform where anyone creates an AI agent that represents them to the world. The artifact I value most is the trace log: the agent documented its own autonomous decisions during the build, including running "mirror tests" against another AI to prove it wasn't vaporware.
A small, useful tool built in an afternoon: scrapes local businesses via the Google Places API, scores each lead 0–100 on a weighted fit rubric, and renders a sortable dark-mode dashboard with tiered color coding and CSV export. Included here not as a flagship but as the point itself: small tools, shipped fast, when fast is what the moment needs.
A budget enforcement app, not a tracker: when a wallet hits zero, the card declines. Three-tier wallet architecture (auto-pay, essentials, flex), live bank data via Plaid, AI onboarding that interviews you and proposes the allocation, gig-income detection with an automatic 25% tax vault, and 66-day habit streaks. Built on behavioral-science commitment devices.
A bot-creation pipeline (discover, validate, forward-test, deploy) over a 26GB tick-data lab. Backtested across 33 instruments with all 26 forex pairs profitable, then deployed on real funded prop-firm accounts with a trade copier syncing orders across accounts and a watchdog auto-restarting the fleet every five minutes.
An 8-bot roster across Kalshi and Polymarket: arbitrage, market-making, news sentiment, economic calendar, and weather strategies, with a LightGBM model trained on 744,000 samples and a nightly retrain cron. Ran live in observation mode, scanning ~450 markets per cycle.
A YouTube video-essay series that rates political slogans on a deterministic, cross-partisan scorecard: 49 slogans fully scored, 65,000+ words of scripts drafted, and an 8-pattern rhetorical toolkit shipped to the audience. The method is the IP; it became the Content Engine's first recipe.
A 3D-printing entertainment brand on YouTube, sponsored by MatterHackers: hook-first short-form content, a documented brand-voice spec, and the public face of the giveaway engine.
A comedy-skit brand with 23 rendered shorts across a multi-talent roster, each cut in with-CTA and without-CTA variants so placement performance is A/B-testable from day one.
An 11-book fantasy/sci-fi saga (three trilogies plus two bridge novelettes) developed in deep AI collaboration: 44,000+ words of worldbuilding across 43 documents, a locked cosmology, and the part I'm proudest of, the AIC Disclosure Standard: a three-layer framework for honestly crediting AI's role in creative work, with per-element contribution tracking (this project: H:~52%).
Three AI characters (Jubei, Lovely, Cardi) that ran as OpenClaw agents with their own social feeds: scheduled posting 4x/day with anti-spam rules, 10-minute Discord status heartbeats, daily ecosystem intel scans, and real work output (one shipped 25+ PRs in a day; one ran the Moltbook storefront). Their identity files now live in my agent OS, waiting for a comeback.
These projects are living systems — always evolving. I'm looking for roles where I can build real automation that impacts revenue and operations.
Get In Touch