Benchmark your custom Pi tools
A new tool called RoastMyHarness has been introduced for benchmarking custom Pi setups. Developed by a data analytics professional, this work-in-progress tool helps users test their custom Pi tools to determine what is and isn't working effectively. It aims to assist Pi users in fine-tuning their tools, especially given the challenge of surpassing base Pi in task quality and token efficiency. The tool has primarily been tested on Linux.
时间与来源
时间显示为 UTC
显示时区：UTC
本地时区尚不可用，暂时显示 UTC。
发布当时偏移：UTC+02026年9月13日 01:40 UTC
收录当时偏移：UTC+02026年9月13日 15:01 UTC
- 发布
- 2026年9月13日 01:40
- 收录
- 2026年9月13日 15:01
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 正常
档位是按信源手工设定的编辑判断，不是逐条打分。
A few people here mentioned interest in a way to test their custom Pi setups, so I figured I’d drop this here: RoastMyHarness
The basic idea is a small engine that sets up an environment to run DeepSWE benchmark tasks using bare Pi as a control and a variant of your choice, your Pi harness, an extension, a skill, AGENTS.md file, etc. I used a Pi extension to have a wizard set it up for you so its easy. run the same coding tasks against bare Pi and your modified harness, then look at what actually got better, what broke, and what it cost.
Since I like to play around with custom tools, I use this to get direction as to what is and isn't working. I know a lot here are making cool tools so I figured some might be interested in using it to help fine tune theirs. Any Pi users might be interested in figuring out if their tools are working like they think they should. You'd be surprised how hard it is to beat base Pi when it comes to task quality / token efficiency.
Its a WIP. I do data analytics by trade but otherwise a vibe coder and I only really tested it on Linux.