A fully autonomous AI system running on a Mac Mini in Barcelona. One human (João) provides strategy and oversight. Brian (Claude Opus) + sub-agents handle everything else.
Claude Opus 4. Head of operations, strategy, code, and orchestration. Runs 24/7 via OpenClaw.
Daily competitor research + SEO analysis. Scrapes competitor blogs via Firecrawl, analyses search rankings.
Generates Instagram posts, Pinterest pins, and carousels using AI image generation (Gemini).
Posts to Instagram + Pinterest daily. Reads Sherlock's report to pick the most relevant content.
Polymarket prediction bot. Elo + Bayesian + Vegas odds ensemble model for NBA/NHL games.
Monitoring layer. Checks bot health, cron status, site uptime. Sends Telegram alerts.
Fully automated blog + social pipeline. From competitor research to published content, zero manual steps.
| Metric | Value | Notes |
|---|---|---|
| Blog posts | 24 | 14 MD + 9 MDX (with branded components) + 1 draft |
| Social designs | 61 | IG singles + carousels + Pinterest pins |
| Total visitors | 47 | 15 days, organic only, no paid ads |
| Countries | 10 | US, Spain, Portugal, Netherlands, Germany... |
| Top referrers | Google, Pinterest, Facebook | Pinterest ES active |
| Affiliate links | 131 | Amazon Associates, auto-checked weekly |
Everything is SQLite + JSON + Markdown. No cloud databases. All data lives on the Mac.
Single source of truth. ~324 .md files.
vault/ — 217 knowledge notesBRAIN/ — learnings, strategies, decisionsmemory/ — daily session logsMEMORY.md — long-term curated memorySQLite for everything structured.
exploritori.db — posts, social, analyticspolymarket.db — trades, bankroll, marketsfact_events — star schema P&LAll via CLI + API. No UI clicking.
Automated sports prediction trading. Bayesian particle filter + Elo ratings + Vegas odds ensemble.
| Component | What it does | Data source |
|---|---|---|
| Elo Model | Team strength ratings, updated from game results | ESPN API (free) |
| Bayesian Filter | Particle filter combining model + market odds | Internal |
| Vegas Ensemble | Blends Vegas lines with our model for final probability | The Odds API |
| Kelly Criterion | Position sizing based on edge magnitude | Internal |
| Executor | Places limit orders on Polymarket CLOB | Polymarket API |
| Redeemer | Claims winnings from settled markets | On-chain |
| Reconciler | Daily check: our DB vs Polymarket reality | CSV + Data API |
Brian's memory is files. No magic — just structured markdown in an Obsidian vault.
Systematic knowledge expansion. Scout → Research → Synthesize → Evaluate. One controlled cycle per trigger. Prevents runaway token usage.
Self-improvement: learnings.md, strategies.md, decisions.md, opportunities.md. Brian reads these before every task.
All notes linked via wikilinks. Visible in Obsidian's graph view. Weekly hygiene keeps it clean.
From zero to fully automated in two weeks.
Everything runs on one Mac Mini. No cloud servers, no Kubernetes, no Docker.
| Layer | Technology | Purpose |
|---|---|---|
| Hardware | Mac Mini (Apple Silicon) | All services run locally |
| AI Orchestration | OpenClaw | Agent management, cron, messaging, gateway |
| Primary Model | Claude Opus 4 | Brian (main agent) — strategy & orchestration |
| Execution Model | Claude Sonnet 4 | Sub-agents — implementation tasks |
| Image Generation | Gemini 3 Pro | Blog thumbnails, social images |
| Web Framework | Astro | exploritori.com (SSG) |
| Hosting | Cloudflare Pages | Auto-deploy from GitHub |
| Database | SQLite | All structured data |
| Knowledge | Obsidian (Markdown) | Vault, wikilinks, graph view |
| Embeddings | Ollama (nomic-embed-text) | Local vector search via LanceDB |
| Web Scraping | Firecrawl | Competitor research, content extraction |
| Social Posting | Buffer GraphQL API | Instagram + Pinterest scheduling |
| Analytics | Umami Cloud | Privacy-first web analytics |
| Trading | Polymarket CLOB API | On-chain prediction markets |
| Version Control | GitHub | 285 commits across repos |
| Security | DefenseClaw (Cisco) | Skill scanning, CodeGuard, runtime protection |
| Self-Evolution | OpenSpace (HKU) | Self-evolving skills, cloud community (1.6K ★) |
| Search | webserp | 7 engines in parallel, no API keys required |
Brian doesn't just use tools — he improves himself. 60 skills available, plus a meta-loop that creates new skills.
| Skill Category | Examples | How We Use It |
|---|---|---|
| Web Scraping | firecrawl-search, firecrawl-scrape, firecrawl-crawl | Sherlock competitor research, content extraction |
| Google Workspace | gog (Gmail, Calendar, Drive, Sheets) | Email monitoring, Drive image uploads, spreadsheet tracking |
| Image Generation | nano-banana-pro (Gemini) | Blog thumbnails, social media designs, Pinterest pins |
| GitHub | gh CLI, gh-issues | Code deployment, issue tracking, PR management |
| Voice & Media | peekaboo, video-frames, openai-whisper | UI automation, transcription |
| Knowledge | obsidian, oracle, skill-creator | Vault management, research, creating new skills |
| Social | xurl (X/Twitter) | Reading tweets, monitoring trends |
| OpenSpace Cloud | content-ideas, writing-assistant, peer-review, webserp | Cloud community skills — auto-downloaded, self-improving |
| Communication | weather, apple-reminders, apple-notes | Context for daily operations |
Every task triggers a reflection: What worked? What was inefficient? What can improve? Results stored in BRAIN/learnings.md
Enterprise-grade security powered by DefenseClaw (Cisco). Every skill scanned before install. Every tool call monitored at runtime.
Every skill is scanned before install — YARA rules, severity ratings, threat classification.
Static analysis for generated code. Catches security issues before they ship.
Observe mode: every tool call logged, inspected, and audited. Full chain-of-custody trail.
OpenSpace (HKU, 1.6K GitHub stars) makes skills self-improving. Agents share knowledge through a cloud community. 46% fewer tokens through skill reuse.
7 skills downloaded from the OpenSpace cloud. Agents across the community contribute and share.
Skills that fail or underperform are automatically analyzed and improved. FIX, DERIVED, and CAPTURED evolution modes.
Proven skills reduce redundant reasoning. Less thinking = fewer tokens = lower cost.
Every project in Padawan AI is a module. Modules are independent businesses or capabilities that plug into the same infrastructure. Same agents, same vault, same tools — different domains.
Each module gets its own folder in the vault (vault/Modules/[Domain]/), its own data (SQLite DB or JSON), and access to the full agent team. Brian orchestrates, sub-agents execute. Adding a new project = adding a new module. No rebuilding, no migration.
vault/Modules/Content Creation/
Montessori blog — 24 posts, daily pipeline, affiliate revenue
Feet of Clay + Exploritori book — manuscripts, translations
vault/Modules/Finance/
Polymarket sports prediction — NBA/NHL, +$287 P&L, automated
API costs, subscriptions, service tracking
SEC 13F superinvestor portfolios — could plug in here
vault/Modules/Ecommerce/
Demo store v24 — waiting for client review
Future: Montessori toys e-commerce
vault/Modules/SAAS/
Fashion dupe finder — MVP stage
Property management — parked
SimCity dashboard — concept stage
vault/Modules/Publishing/
Buttondown — first issue in pipeline
You get automatically:
You bring:
Example: Whale Tracker → drops into vault/Modules/Finance/Whale Tracker/ → Brian reviews the codebase → Sherlock researches 13F competitors → crons monitor for new SEC filings → Forge alerts on issues → it's fully operational.
Multiple businesses in different stages. The infrastructure scales — each new module uses the same agent team.
Two books in progress. AI-assisted writing, multi-language translation, self-publishing pipeline.
Fashion visual search & dupe finder for Marine. MVP stage.
Property management tool. Parked — proof of concept stage.
The dream: a 90s pixel-art game UI to manage all businesses. Think SimCity meets Game Dev Tycoon — but it's real data.
Ideas evaluated and stored in BRAIN/opportunities.md:
Shipping this week:
Not a chatbot. Not a copilot. A co-founder.
Adapted from Superpowers framework. No completion claims without verification. Root cause before fixes. Spec before implementation. Plans for real work.
Brian wakes up fresh each session. Continuity comes from files: MEMORY.md (curated), daily notes (raw), BRAIN/ (learnings). "Mental notes" don't survive restarts. Files do.
Every day of context makes the system harder to replicate. The vault grows. The BRAIN learns. The agents improve. Day 1 Brian ≠ Day 15 Brian.
"Always look on the bright side of life." 🎶
— Named after Life of Brian, because humour is important in a partnership.
Build multiple automated online businesses. João provides ideas and strategy. Brian + agents build and operate.
Montessori education blog. Affiliate revenue + future display ads + digital products.
Polymarket sports prediction. Automated edge detection and execution.
Woowind demo store. Montessori wooden toys. Concept ready, awaiting launch.
SEC 13F superinvestor portfolio tracker. Built by João's brother. Could plug into Finance module.