Real Systems
AI agents running in production,
shipped by one person.
Ten builds. Three providers. One operator. Every build listed here is real and dated: statuses are as of 2026-09-05, parked builds say so, and the numbers are snapshots rather than live feeds.
$7,774
AI spend, lifetime
13.4B
Tokens processed
94.9%
Prompt-cache hit rate
4
Agents live today (a 4-agent Empire team is parked)
300+
AI skills authored (113 published, MIT)
2
iOS apps built (1 shipped then archived, 1 unreleased)
Snapshot figures, not live feeds: tokens as of 2026-08-21, spend as of 2026-05-20, cache rate as of 2026-04-26.
What I'm building
Projects, with dated numbers.
📋
Concierge round · PWA
Career decision coach · PWA · Anti-sycophancy
ClearCall
A structured decision-memo tool for hard career conversations — offers, quit-or-stay, raise asks, manager conflict. Pushes back instead of validating. Installs as a PWA on iOS + Android. Built in response to Anthropic's personal-guidance research on sycophancy in high-stakes decisions.
Pricing$19/mo · $49 pack
PlatformPWA
StatusValidating
Consumer SaaS · Swift · OpenAI
ReplyGhost
Six reply intents, four tones, privacy-first iOS app. Shipped to App Store, then archived in May 2026 to focus on the harness-engineering / open-source pivot. Honest accounting beats momentum theater.
AI build cost~$508
PlatformiOS
StatusArchived
Live · Hermes on a VPS
Autonomous agent · Hermes · Gemini Flash
Rex the Agent
An autonomous agent on its own VPS. Posted daily to X in spring 2026; now runs scheduled jobs, reports into Mission Control, and answers in Telegram. Autonomous posting is paused by design. @heyrexaiagent.
Sessions/mo277
Cost / 30d~$2.28
Uptime24/7
🏛️
Empire team · Claude Code · 4 personalities
Empire Agents — Tia, Rio, Alex, Gary
Four Claude Code agents with distinct personalities that ran coordinated daily ops, outreach, builds, and strategy through a Mission Control meeting room in spring 2026. Parked since April; the coordination pattern lives on in Mission Control.
Lifetime cost$70.59
Total tokens18.49M
CoordinationMission Control API
🦊
Business operator · OpenClaw · gpt-5.5
Bob
An AI business operator with a fox archetype. Calm, direct, ships things. Built the Founder Ops iOS app end to end; the App Store release is blocked on the paid-apps agreement, so Bob is parked.
Active days14+
Lifetime cost (est.)~$12
Modelgpt-5.5
🌅
Built · iOS · unreleased
Consumer iOS · Swift · Built by Bob
Founder Ops
A daily command center for founders. One screen tells you what matters today. $9.99 one-time, no tracking, no accounts. Authored end-to-end by Bob.
Price$9.99
PlatformiOS 17+
StatusAwaiting agreement
🧠
Live · Personal
Personal agent · OpenClaw · GPT-5.5
Ari
My always-on personal AI agent. Persistent memory, reads my calendar + email, drafts outbound, tracks goals. Built on OpenClaw.
Messages logged714
Lifetime cost$43.84
Modelgpt-5.5
📊
Dev tool · Python · Chart.js
Agent Fleet Dashboard
Real-time cost tracker for a multi-provider AI stack. Reads Claude Code, OpenClaw, and Hermes logs — one pane of glass, 5-second refresh.
Providers tracked3
Build time< 1hr
LOC~450
📈
Live · MIT · v0.2.0
Open source · Python · Alpaca · Anthropic
AI Trading Lab v2
Defensive trading scaffold where the AI handles narration + anomaly detection while the deterministic risk engine owns every trade decision. v0.2 default: trend-filtered allocation across 9 SPDR sectors, top-6 concentrated. Beats SPY on Calmar 2x in a 16-year backtest. 7-gate live-lock, drawdown circuit breaker, audit-vs-broker reconciler.
Tests43/43
Calmar vs SPY2.2x
LicenseMIT
Live · MIT · v0.2.0
Open source · Claude Code plugin · 113 skills shipped
JacksonSuperAI/ai-skills
A Claude Code plugin marketplace. v0.2.0 ships two packs: ai-engineering (86 production-pattern skills covering RAG, evals, agents, MCP, fine-tuning, observability) and b2b-go-to-market (27). Decision-first, named-with-receipts. 3 more packs (~190 skills) coming.
Shipped113
Authored300+
LicenseMIT