Quizzical

TL/DR - my current agentic programming model is to use Claude Code with another machine running qwen. Both machines use LLMs and agents to work on one task sequentially on one clone of a codebase. Sometimes I use multiple clones and accept git diffs from remote agents handed tasks. Sometimes well known companies produce questionable code using AI that you are running and probably don’t know about. I’m writing this with a slight cold and a touch of fatigue, so if that comes through, please excuse it. ...

As Napoleon Dynamite said … This is a collection of skills I have been evolving for my own project. I typically use my Claude Pro subscription in a terminal on my Framework and, to preserve tokens, assign grunt tasks to qwen3-coder:30b running on an RTX 4090 on a gaming machine I bought from Scorptec two years ago. My primary motivations for building out these skills have been: ...

When DHH made his ‘Rails is Omakase’ statement last year, Tom Dale tweeted: Rod Vagg pointed this statement out to me in February, while we were talking about the lack of conventions in the Node world and whether it was a good or bad thing. In the Ruby world there is a ‘Rails Way’, lately nearing authoritarianism (although I know of Ruby devs who won’t do Rails). In the Node world there isn’t an equivalent option. ...