Building TM Power Workstation: a single-developer agent pipeline
Why I stopped trying to hire a team and built one out of LLMs instead — and the dependency-graph engine that makes it work.
Direct LLM inference vs. Bedrock: when each one wins
A field report from running production workloads against both. Cost, latency, and the subtle differences in how prompts behave across providers.
Multi-tenant storefront rendering on a single Spring Boot host
How the StoreBounty engine serves dozens of domains from one JVM — and the request-pipeline trade-offs that made it possible.