為了找到一些在地端也能讓 Agent 有無限 token 自由的毒駕的方法，原本用手邊的M4 24GB Mac 上嘗試執行 DiffusionGemma 26B，卻悲慘的連 1,000 tokens 的...
DiffusionGemma 26B 登陸 M2 Max：MLX 吞吐量實測與 Context 極限挑戰
Related
How I built a mind map that's just a Markdown list (and why that makes AI streaming almost free)
A technical breakdown of Open MindMap: turning an indented Markdown list into a live SVG mind map, and why that single design choice makes LLM streaming fall out for free.
I built a Claude Code skill that finds customers, not competitors, on Reddit & LinkedIn
`I do consulting. Like most people who sell a service, I know I should be active on Reddit and...
AI Agent Evaluation Harness: Test Real Workflows Before Users Do
Build an AI agent evaluation harness with task fixtures, trace scoring, judge checks, regression tests, budgets, and human review before agents fail in production.