Yup, I made my own code editor. My goal is to make a simpler editor just for css, html, and markdown. This might be a bit of going down the rabbit hole of making tools to make tools, but there are some specific features I couldn't find elsewhere.

Mainly: View diffs highlighted inline. AI writes big chunks of code all at once, and I want to see what it did. Default is 'uncommitted changes' (which is what Zed gives you), but I also have 'last commit', because sometimes the agent commits all in one go.

This is my favorite tool to date. Modes are saved window layouts under a Project. This spotlight tool pops up on ctrl+space, and allows changing Projects and Modes with a few quick keystrokes. Tab to edit.

This simple geometric design is more of the style I'd like to do for this project. Most everything is tidy and wireframe-y. I hope to push things out of the beige zone in short order.

I’m building my own personal AI operating system.

I’ve worked as a UX designer for emerging tech for more than 10 years. I’ve been fortunate to work on many creative products, from 3D printing to VR and games. Unfortunately, I’ve also seen how the incentives of the tech industry lead to consumer technologies made to extract our attention and lock us in to a corporate platform.

AI is different. I think the best thing about AI is that it has created a turning point where anyone can build technology for themselves. Finally we can make personal personal computing and free our attention from misaligned algorithms and the notifications nudging us to use features we don’t want.

This summer, I’m replacing all my daily consumer apps and devices with self-built versions, and will use and iterate on them over several months. Many will be exceedingly minimal, since I only need to include features that I use. Some will have unusual forms or materials, since I seek to get away from the industrial utility of plastic and metal, and craft forms are more personal and malleable. I will incorporate AI with intention and low distraction. I will document extensively both as an act of self-reflection and to bridge the communities of open-source and DIY.

This is a research project to explore my own mental space and express it with the aid of computing. It’s also a rebellion against a tech culture that espouses domination and scale.

My plan is simple. Whenever I find myself frustrated with doing a computer thing and can imagine it better, I’ll build myself a little app that does exactly what I want, and maybe add a bit of a spin for experimentation. I’ve already made a few. With Claude Code, it’s possible to make a little personal app in less than a day, and that’s exactly the point. It is sometimes faster to build the thing you want rather than searching online and installing different apps from the App Store.

I’ve got a list of ideas but I’m trying not to plan so much. Build in the moment and don’t extrapolate. So far I’m going to keep my Mac and iPhone since I already own those, but I do want to try more open source hardware and seeing how much that can handle.

It feels like it's faster to build the apps than to document building the apps.

Most of these apps I built in a day. Some of the little ones were less than an hour...just one or two prompts.

I procrastinated on making this website because I keep wanting to make the work a bit better before sharing. Even though I always intended to share work in progress as part of this project. Most of the features work, but I would consider the current design to be 'wire-frame'.

I've learned a lot about developing on MacOS. I read most of what Claude writes when it builds a feature, so I maintain a general understanding of how everything works. Architecting is still the challenge, and a good amount of that depends on my preferences for UX. It is possible that I'll re-build just the parts I like after my opinion on how everything interacts together evolves.

So, 3 week benchmark. 12 apps. I'm still on the $20/month Claude subscription. I do my own visual testing and sometimes write some CSS. I don't use subagents, because my code base is fairly simple and I found that overcomplicated tasks. I've learned some tricks: refining the Claude.md file impacts the work a lot more than prompts, and it prefers looking at an example over reading more docs.

Now I want to be refining what I've built instead of generating more apps. Although I also want to test some more weird ideas and ideas more relating to the real world, not just working on my computer. Ironically, working too much is a problem. I have better ideas when I've had a restful weekend, and I want my tools to reflect that.

I noticed people using LLMs to generate slides. Powerpoint is far too heavy for this. People soon figure that out and then start doing Markdown or Html slide decks. There's many advantages for using AI to summarize content and put it into slides.

I feel this needs its own editor. Mine is built on a simple json schema. The LLM can easily write json and focus on editing content. Then use the editor to apply themes and do edits.

My specific POV here is that I want to change themes after the slides are made, and it applies to the whole deck. Not pick the theme first before I get started. I also believe templates should be simple. I welcomed Figma Slides as a huge improvement over Google Slides, but still have found their themes to be over-complicated. And somehow editing and copy/paste still takes way too long in Figma slides. (Try adding links to pages and editing them, it's infuriating.)

Html is great for slides because it has all the text flow and styling features you can ever want, and publishing slides online is innately useful. But I don't want to be editing html and css to make a presentation, even with the help of an AI agent. Themes and templates serve the purpose of helping to structure the presentation without too much distraction from the content.

This is one of my more ambitious ideas. To make a simple tool that facilitates collaboration with AI yet still has a lot of design controls. So far this has been just over 1 day of work, so I haven't put too much effort into making it nice yet.

My first decision was to have a menu bar dropdown list of Projects. I'm always working on multiple things, and I got tired of rustling around in Finder or making bookmarks to folders. Having this little menu is so satisfying.

Underneath it's still files in local folders, because I don't want to upload more things to the cloud. Plus I get all the functionality of existing local files. Deleted items go into the trash bin, and I can restore them if I want.

Maybe the most significant mental sorting difference is that I've made Projects the top hierarchy. Not apps. Honestly I care more about my projects and the ideas they contain than apps.

I've used Github Desktop for a long time. But it only gives you one window, and it doesn't shrink below a certain size. So I made my own Git helper, which is much smaller and simpler. It just has pending commits, the button to "commit all", and the last commit. Clicking a file opens it in the Code Editor.

The bright colors are set per project. It's my favorite feature. I can see what Project I'm in right away, and it's a bright post-it block on my screen.

This was the first tool that I built. Mostly, I wanted a better way to organize all the images in a project folder, and I didn't want to use Lightroom for everything. Finder has thumbnail view, but the thumbnails are too small, and it's difficult to get the view settings exactly right.

The main feature is copy/paste. By default, cmd+c copies the whole image. So I can copy right over to Figma. If the toggle for the Image Editor sidebar is open, cmd+c copies the Edits, and you can paste those Edits onto another image. That's my favorite feature from Lightroom, and it was satisfying to build my own.

This might be my most used Tool. It was the second thing that I made. I was writing in a combo of Apple Notes and Stickies, and that really wasn't working out. I love Apple Notes, but I just needed my own notebook app that could feasibly have all the customization I would ever want.

It was really satisfying making my own Table feature. It's just a super plain table where you can add sum columns. Because that's really the most common formula that I use. It's silly to have a whole spreadsheet when I just need to sum.

I built my own image editor because 90% of the editing I do is super basic. I put images into Figma just because I like their simple sliders and export functions. I hate opening Photoshop. I also use Preview to rotate images and remove backgrounds.

Most of the handiness is in almost invisible: -Remove background using Apple's native, super-fast feature. -Real export for web, including WebP. -Copy/paste all the edits from one image to another. -Loads super fast.

I also tried making an 'extend background' feature, but it doesn't work super well. Usually when I open Photoshop, it's just because I need to expand or clean up the background. I spent too much of an evening trying to get a local image generation model to work. It looks like the next MacOS release will include native background extend, so I'm fine with just waiting for that.

Daily Note lives in the top menu bar with its own icon. I made it because I will make a 'to-do' list, and then several days later make another one. The items are usually low stakes, and there's no point in saving all the different to-do lists. If something is on there for ages, and I still haven't done it, I just delete it. High priority things that I super definitely have to do get a calendar event. Or I actually just remember, because I don't bother remembering the little stuff.

So this is just One List. Any unchecked items get carried forward to the next day. I put the calendar because I think it looks nice at the bottom, a bit like a stationery page.

Information is liquid, and reshapes itself to fit the container. We endlessly reflow knowledge from one container to another in efforts to grasp its truth. Yet the form of the container also affects its function.

A slide deck is not the same as a research paper. A blog post, press release, social media post, all different containers. A website, a video, a book. While the industry is fond of predicting the 'death' of a medium when a new technology for knowledge distribution becomes available, the truth is that humans thirsty for insight and inspiration will gladly add more containers than subtract.

We build new containers when there is the feeling of missing something. Missing on the idea or feeling or grand concept of the author. Classes, lectures, webinars are containers. So too are comics, zines, movies, music, games.

An illustration is beautiful not because it looks nice next to an article, but because it adds a lens for seeing. A soundtrack doesn't fill silence, it creates an entire mood and space to inhabit.

LLMs are the container we have crafted that is closest to the native liquid nature of raw information. Its flow is innately a firehose. Give it some containers and it will gladly reflow.

A single column file browser that lists all my linked folders in a Project. Because a Project might often have a code repo and a folder of images or videos that aren't in the main Project folder. It lives in its own window so that I can make it extra tall. A super bright color highlight, toggle for font size, and single click to open.

This is a second iteration of a super simple idea. I have 24 GB of RAM on my MacBook Pro. That should be plenty, but somehow I would get to a point where everything is lagging like crazy. I want to this info at a glance, without opening Activity Monitor. If the number gets close to 20, I click it to see what's hogging the memory. It's usually extra browser tabs, or a dev server that has been left running.

It sits in my top menu bar and expands if I click it. I had looked for an existing app that does this, but couldn't find anything that's this simple. I don't need a meter of everything, I just need to know when to close some apps.

Every interface now feels outdated. I notice the effort I put into dragging windows on my screen. Finding the application after it’s minimized. Searching for files. Searching for the right app. Searching for the settings panel to toggle the right setting. Searching the web to find that settings panel. Settings is the junk drawer of any app or OS, and yet it’s the drawer I most frequently open.

Even for apps I use regularly, there is no interface where I do not have to search to do the thing I want. The apps are always updating. The familiar button I have pressed for years is suddenly somewhere else entirely. There is ever more clutter of new features, on surfaces where I only use 10% of the current features.

I do not need more screen time. And yet every software experience seems to be designed to demand my time and my eyes.

I also don’t want to talk to the computer. At times it would be useful…when I’m driving, or cooking…the computer should be able to deliver on straightforward requests. But the car dash doesn’t see the text message for changing the meetup location. The Echo doesn’t know the recipe I’ve loaded on my phone. And they talk too much…paragraphs when I expect 10 words.

I love taking a walk without my phone. But then I want to take a photo of the clouds overlooking the ridge. I love drawing in my sketchbook. But then I want to undo that stray mark. I love seeing my friends in person. But many of them live thousands of miles away.

Technology eagerly shouts, “I am here for all of those needs!” Yet the experience is unsatisfying. A meal delivered that is too many empty calories.

Notifications notifications notifications. I’ve paid thousands of dollars for pieces of aluminum and glass that seem entirely focused on delivering notifications that I don’t care about. I turn all of them off. But then I miss my messages. How do I just get the messages. So many different settings. Yet impossible to achieve this simplest of intents.

I’m scrolling again. Why? I could be doing anything else. But I want to know what’s going on in the world. What are other people doing? I scroll. This video is so funny. Look at this cute thing! Wow this new product is impressive. This person is upset. No it’s just rage bait. Here’s a useful post. Bookmark. Now where did I save that to? Oh no I’ve been scrolling for 40 minutes.

I’m logging in to an account for something important. What did I put for these security questions? Mother’s maiden name, high school, pet’s name. But I didn’t put the real answers of course. What imaginary life did I conceive of when creating this account? I made a super secure password for this site. But now I can’t autocomplete it…I have to type it in. So I copy and paste this obscure incantation of characters into Notes, so that I can see it to type in on my device. I shouldn't admit that I do that.

There’s a storm and the power is out. The internet is out. Why do I sit down at my computer and open the browser? What is it that I need to do? Muscle memory that supersedes utility.

It's not that interfaces look ugly. Quite the contrary, many things look beautifully organized. Yet beneath that tidy surface, our technology has been built around behaviors of distraction and anxiety. And it unabashedly shoves this into my life as a baseline expectation. It's an artifact of consumer culture that feels more and more misaligned with the life I seek.

A punchcard of my git commits.

I made my own UI for Claude Code. I put cute animated animal sprites, custom set per project. It has tabs, which the Claude desktop app doesn't have.

It doesn't quite work well enough for me to use it reliably yet. But it was fun and satisfying to get this far in a day.

The simplest thing. It prints the window size of whatever window I click on. Yet so useful!