Lately my interests revolve around good design, minimalist UIs, and making complex workflows easy and intuitive. I've experienced time and time again that we can still unlock huge productivity gains with better UX.
I believe that if there is any point to software, it's to augment human capabilities and provide new tools for thinking. See interesting problems below for some examples.
Here is some of what I've built in the past:
I often come across problems for which I'm not aware of an existing solution. This is basically a list of projects I'd like to work on.
Reach out if you have any insights or ideas, if you want to collaborate, or if there are existing tools/products that I have overlooked.
Tutorials and blog posts should more often be centered around letting me see the diff of the code in a particular step. Currently a common dilemma for authors is that they have to choose between:
Either way the result for the reader is often a lot of jumping back and forth, trying to piece together a full picture of what's happening. Why can't I just jump through the diff while reading the article next to it?
One possible solution: use a git commit history to render an interactive diff viewer next to the article text. The diffs could advance automatically when I scroll to the next section, and I want to jump to the full source code view at any time, so that I can explore the changes and the full context at my leisure.
Side note 1: It would be fun to try the same thing for teaching writing or other topics centered around text.
Side note 2: This exposes another belief I have: Despite it's unquestionable success, version control is still not used to its full potential, especially outside of software engineering.
Possible problems with my approach:
I want an editor plugin that lets me write SQL queries against files with tabular data (mostly csv, parquet, etc).
I want to see the resulting data right there in the editor. I know of Rainbow CSV, but I think the experience can be improved. The query engine is awkward, and I have to address the columns by index instead of by name.
The approach I would like to try is to dump the data into sqlite and do some simple detection of column types. Then simply send the query to sqlite and make the output pretty.
Prior art: q is a tool that lets you do exactly this from the commandline. This is great for a lot of use cases, but I also want editor integration so that I get syntax highlighting and a more interactive experience.
I like making furniture. This is where I will gradually put plans, and photos of the more interesting projects.