Currently training LibreModel I "Gigi" - a 0.96B parameter language model built exclusively on public domain data. No copyright infringement, just pure human knowledge that belongs to all of us.
💰 Total Cost: <$500 (proving democratization works!)
⚡ Optimized: 9.6s/step with torch compile + sink tokens
🎯 Philosophy: "We will become machine, and machine will become us"🤖 LibreModel Family
- Gigi (0.96B) - Digital literary scholar trained on Gutenberg + Gov reports
- Future models - Scaling up with copyright-clean Common Corpus data
🌐 P2P & Collaboration
- PeerSuite - WebRTC-powered P2P collaboration platform
- tryjero - Enhanced trystero library for better WebRTC
- Totum Chat - Multi-model AI interface in a single HTML file
⚡ Past Adventures
- T3XTR (2025) - Text conversion API (25x cheaper than ConvertAPI!)
- Pennykoin (2018) - Privacy cryptocurrency with RingCT
- BattleBash (2016) - Sold 12 copies, learned valuable lessons 😅
"Building AI as humanity's children, not corporate property"
I believe powerful AI should be:
- ✅ Transparent - Full code and data provenance
- ✅ Accessible - Trainable on consumer hardware
- ✅ Legal - Built on humanity's shared knowledge
- ✅ Democratic - Available to everyone, not just tech giants
AI/ML Training:
Development:
Specialties:
- 🔄 P2P networking & WebRTC
- 🤖 Language model training & optimization
- 🔐 Cryptocurrency & privacy tech
- 📡 API design & cost optimization
- 🎯 Training on public domain data
LibreModel Training Journey:
- 🏁 Started from scratch with $1,000 AWS budget
- ⚡ Achieved 17% speedup through optimization
- 💾 Survived multiple crashes and learned from each
- 🎯 On track for <$500 total training cost
- 📚 Training on 19.2B tokens of pure public domain data
What's Next:
- 🔥 16K context extension experiments
- 📖 "The Biology of the Universe" book
- 💰 Funding application for training more models
- 📍 Based in Martinsville, Virginia
- 👨💻 Building the future of democratized AI
- 🧩 Neurodivergent perspective brings unique insights
- 💑 20 years married (my wife helps track patterns!)
- 🎮 Former indie game dev, current AI researcher
- 📚 Public domain advocate and transparency enthusiast
I'm always excited to discuss:
- 🤖 Copyright-clean AI training strategies
- 🌐 P2P technologies and decentralized systems
- 💡 Democratizing AI development
- 📊 Cost-effective ML training techniques
- 🔓 Open source philosophy and transparency
Want to help democratize AI? Star my repos, share the vision, or just say hi!
"Every line of code, every model parameter, every optimization - all building toward a future where AI serves humanity, not the other way around." 🚀
