I wanted to run some tests with GLM-5.2 and opencode stalled out on me twice, so I quickly implemented pi agent and hung it into my framework (w/ ppq) and this seems to work better than opencode. It finished the assigned job without issues and with minimal token burn. I'll re-test Qwen 3.6 with this later in the week / weekend.

A Sovereign Brain on a Laptop: Local LLM + Pi Agent + Markdown

rolznz

Reporting back in.

> Did you try Pi Agent?

I wanted to run some tests with GLM-5.2 and opencode stalled out on me twice, so I quickly implemented pi agent and hung it into my framework (w/ ppq) and this seems to work better than opencode. It finished the assigned job without issues and with minimal token burn. I'll re-test Qwen 3.6 with this later in the week / weekend.