Cozy layout

First example of a “vertical cliff” that came to mind: data scientists who want to take some script and run it A) at 3 am, or B) on a machine with more memory. They have to jump from RStudio on their laptops to cloud accounts, containerization, etc.

Do you think there is something that is structurally difficult about building a Notion or an Airtable? In the sense of economics or even something like cultural values? People say it’s differentiated on design, but I suspect that there are many people who could design/build products like that, and there’s some other, invisible hurdle that those companies overcame.

Great read!

I saw patchwork at local-first 2025 and was super impressed.

This is really well done. It illuminates the motivation, explores some of the objectives, and doesn't shy away from unanswered questions and newly created problems, still making a strong argument to keep exploring.

This is really cool. In my experience, the best tool I've got so far is Obsidian - the data layer is just a file system so I could build upon it - stuff like making my own sync layer (tossing the Obsidian vault into a folder that syncs across devices) or using it as a CMS (sourcing posts from .md files). I've even been using Obsidian itself as a data layer for some personal projects. There might be a lot of potential here with stuff like Bases and such. Of course, Obsidian is singleplayer, which rules it out for a lot of things.

I've got a few potentially relevant observations.

I've seen people abandon structured tools (Apple Reminders, Trello) for completely free-form things (tldraw, Apple Freeform), removing any structural limitations. That makes me really interested in I&S's Embark - "gradual structure" type of thing, where you start with completely freeform data and gradually enhance it when you figure out the structure you want. I've personally adopted an approach like this for note-taking: ezhik.jp/hypertext-maximalism

My final observation comes from experience. At work, I built a tool for doing quality checks on data. I started with a static set of checks - but ended up fielding a lot of requests to make little adjustments to the checks. I added more and more knobs and toggles, eventually hitting the "oops, you are building a programming language" problem. Eventually somebody had a request that would require adding loops, and I've taken a completely different approach - I tossed my existing configuration functionality out completely, and instead gave the users access to the underlying SQL engine. And people absolutely love it. The barrier to entry is low enough as most users have at least a basic understanding of SQL. Plus, this is considered a "low risk" tool due to the safeguards I've included, as opposed to more "proper" scripting languages. And so people started to build and build and build.

That makes me firmly believe that lowering the barrier to entry for coding is really the way to go - more "bicycles" and less "aircraft carriers", to borrow a phrase from P.v.H. I think lowering the barrier to entry is super important. From experience, I believe that it's much easier to teach somebody JavaScript or Python essentials than it is to teach them to set up npm or pip or git or vite or...

It's why I personally love user scripts in web browsers - I can write some code without leaving the browser, as opposed to a browser extension that requires an IDE, a build process, and approval by the browser extension store. We really lack the toolset for making (not to mention distributing) "home-cooked apps", which is why that Discord screenshot of Geoffrey Litt sharing a custom Patchwork tool on Discord means the world to me. It would be so cool to see more tools and workflows like that.

Hard to avoid mentioning AI here - but I do think there's good potential here too. My best experiences with AI tools were when I used the AI to build the tools - essentially, asking it to translate English to code. I'm imagining a Patchwork future where the tools can be worked on in Patchwork itself. Both humans and AI agents working on them...

@Ezhik Thanks for sharing all those thoughts and anecdotes. On the point of AI, we've done some experiments with a patchwork tool that's an AI chat interface specialized for making new patchwork tools. Mixed success (this was last fall, FWIW) but yeah, it felt promising.

Hope you'll share more about Patchwork extensions in the future!

This is a bit of a tangent, but this post created a somewhat heated and extremely interesting discussion over in the Malleable Systems forum. It also connected up for me with my hobby horse of "problems" that I ran with here at FoC last year.

Stifling of Imagination open.substack.com/pub/programmingsimplicity/p/stifling-of-imagination?r=1egdky&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Paul Tarvydas I'm curious: does your post feel like more of a response to the top-level post or my link right above? Feel free to say a ratio instead.

Kartik Agaram top-level