I Pushed Local LLMs Harder. Here's What Two Models Actually Did.

In Part 1 of this series, I set up Claude Code against local LLMs on dual MI60 GPUs and watched it scaffold a Flask application from scratch. Small tasks worked. Complex ones did not. I ended with three ideas I wanted to test: running a dense model, trying Claude Code's agent teams feature, and building a persistent memory layer for coding sessions. I started the experiment that mattered most: giving a local LLM a project with real scope and seeing what happened. I ran the same project against two different models. The results were instructive, and not in the direction I expected. The Project The model's goal was to build a Python CLI tool called health-correlate. It would connect to my InfluxDB health database, retrieve time-series data for metrics like glucose readings, blood ketone levels, blood pressure, body weight, and subjective wellbeing scores, resample everything to daily aggregates, and run Pearson correlation analysis with configurable time-lag support. The output: a ranked