The tool-switching itch cost me a year. Every new model felt like the answer, so I never got past beginner with any one of them. The week I froze my stack and just deepened my CLAUDE.md, output doubled. Depth beats novelty. And yeah, mine was 300 lines the agent never actually read.
You’re not alone. That is what everyone does in chasing the new model of the week. I did the same with several projects in the last few months, cleaning, purging, and restructuring to “lazy-load” only what the model needs at that very moment. This keeps the context clean at every turn.
Thanks Sandy - this seems like a bug in the Substack app. Copy seems to work in the browser in mobile and desktop UI. I just submitted this as a bug to Substack.
The tool-switching itch cost me a year. Every new model felt like the answer, so I never got past beginner with any one of them. The week I froze my stack and just deepened my CLAUDE.md, output doubled. Depth beats novelty. And yeah, mine was 300 lines the agent never actually read.
You’re not alone. That is what everyone does in chasing the new model of the week. I did the same with several projects in the last few months, cleaning, purging, and restructuring to “lazy-load” only what the model needs at that very moment. This keeps the context clean at every turn.
Good article for me, a beginner. But the paste boxes do not copy.
Thanks Sandy - this seems like a bug in the Substack app. Copy seems to work in the browser in mobile and desktop UI. I just submitted this as a bug to Substack.