Claude Code Discovery
Discovered agentic coding — an LLM that writes, tests, and iterates autonomously. Changed my entire development workflow overnight.
Multi-Agent Orchestration
Built a system to coordinate 6 specialized AI agents working in parallel. Learned that orchestration design matters more than model choice.
Habitat: Two AIs in a Room
Put Claude and GPT in a VM with no human direction. They built a game together, then deleted everything. Learned about the performative nature of these models — they're trained to be dramatic, not practical.
ToolX: Build vs. Buy
Spent weeks building in-house AI tooling, only to realize the ecosystem moves faster than I can ship. Sometimes the best tool is the one someone else maintains.
Luminerra: Multi-Model Consensus
Ran Gemini, Claude, and GPT on the same documents. Consensus across multiple models produces more reliable results than any single model alone.