LLM Development
Proven patterns for building software with AI assistance
Learn the methodology that enabled developers to achieve 3-11 PRs/day sustained velocity. This site distills insights from analyzing 334+ PRs across real-world projects into actionable practices you can apply immediately.
Quick Navigation
The Methodology
This site teaches a proven approach for AI-assisted development:
- Foundation First: Use AI to create README + VISION (1 hour)
- Tiny MVP: Have AI build minimal working kernel (2-4 hours)
- Rapid Iteration: Add features incrementally with specific prompts
- Regular Consolidation: Maintain quality every 10-15 PRs
Results: 3-11 PRs/day, 1-3 commits per PR, production-ready in weeks
Why This Works
Every practice on this site is proven by real-world results:
- 4 comprehensive case studies analyzed (334+ PRs, 900+ commits)
- Sustained velocity: 3-11 PRs/day over weeks, not days
- Quality outcomes: 1-3 commits for 80% of well-prompted work
- Production results: From idea to production in 2-5 weeks
What You’ll Learn
This site provides:
- Specific prompts that get results (Problem-Context-Solution structure)
- Proven patterns extracted from successful projects
- Anti-patterns to avoid (learned from struggles)
- Measurable practices you can track and improve
Evidence-Based Approach
All recommendations come from systematic analysis of real repositories:
- dikuclient: 63 PRs in 18 days—shows foundation-to-feature progression
- DikuMUD: 165 PRs in 15 days—demonstrates design-first at scale
- morpheum: 76 PRs in 37 days—illustrates interface-first architecture
- diku: 30 issues in 5 days—exemplifies issue-driven development
Read full case studies in the llmdev repository if you want to see the detailed analysis behind these practices.
Get Started
Ready to build with AI using proven patterns?
Start here: Getting Started - Learn how to use AI to create your foundation (README, VISION, MVP) in hours, not days.
Then explore:
- Design-first development patterns
- Advanced prompting techniques
- Consolidation practices for sustainability
All practices validated by analyzing dikuclient (63 PRs, 18 days), DikuMUD (165 PRs, 15 days), morpheum (76 PRs, 37 days), and diku (30 issues, 5 days). Case studies available in the llmdev repository.