Architecture
5 min read
Architecture
Multi-Agent Systems vs. One Good Agent: What Two Years of Agent Development Actually Taught Me
I built nine agents for one feature before I noticed eight of them were passing notes to each other about a job a single agent could have finished alone. A planner. A researcher.
→
Llm Infrastructure
5 min read
Llm Infrastructure
The Sixth Lesson of Local-LLM Orchestration: Why Your Harness Decides Your Output, Not Your Model
I ran the same multi-agent research task through a 70B model and a 14B model on the same box last spring. The 14B won. Not on speed — on the actual answer.
→