adapting your tools

One of my favourite questions to ask people who recommend Claude Code or other AI tools is: what did you make? If you were building interesting things before you will probably be making interesting things now. However, you might also get stuck in optimization loops or trying to have 6+ windows with agents going all the time but not actually creating anything.

One way to use these tools is to append and improve existing tools. I tried an early experiment on trying to create book summaries, research or find novel connections within my notes. All the results from these experiments were pretty lacklustre. Then I simply used it to create tables to collect undone tasks from separate project notes across my notes (in obsidian, through a plugin called data view that uses javascript).

I was following along video tutorials but the code was not quite as how I wanted. This vastly improved the design and usefulness of Obsidian, compared to the initial or commonly spoken about use cases.

Similarly, with this website. I didn’t like Wordpress and Substack. I just wanted someplace to write without friction and a clean design, focused completely on reader experience Inspired by the London Review of Books and Edward Tufte style footnotes on the side. It took an evening and turned out exactly how I wanted it.

For me, the best part of coding agents has been being able to adapt my tools to myself. Not having to adapt to shitty, poorly made mass market tools. Software companies get acquired and tools disappear. The extra effort to build something bespoke or build on top of something to make it more personal. there are still very many tools that are not worth creating or augmenting that work really well out of the box and are worth paying for.