I have finally learnt to do fewer things.
Time feels scarcer, and I was forced to “prioritise”. My idea list continues to grow but I prune it as often as I add to it. It’s slightly lame to be limited to software, but certain projects must be completed before I move on to other things.
A full rewrite. Adding 4 more clients (E-mail, Slack, API, Raycast), a better web experience, and more customizability for Stream designs.
I’m calling it a “UI builder”, working off of the idea that most of the apps you see are made up off the same primitives (documents, files, whiteboards), with only marginal layout/styling differences.
The editor I have in mind lets you control the layout first, and then add in each type of object you want to work with. Publishing, customizability, and offline editing will be first-class features.
A collaboration with Abhimanyu. Not much to say about this one until it’s ready, except that I think I did a good job on the design.
If you like the things I make, you can pay me for them, or hire me to make something.
I have no deadlines, I just ship as much as I can in order to stave off the unrelenting ennui. This list is for personal reference more than anything else.
Cozysearch – Semantic search over a manually-curated corpus of webpages.
Pro Tips – A guest-authored blog.
Static sites: index.html | Hugo
ORM: Drizzle
Hosting: Cloudflare | Digital Ocean | Netlify
DAW: Garageband | Logic Pro
Design: This sketchbook | tldraw
We need a Wordpress competitor that's made for static sites.
Think about formats. Think about "posts", and "collections", and "updates". Think about creative freedom and power alongside simplicity.
Needs to sync to a Git repo (or does it??), needs to provide live previews while editing, needs to add things like search, dynamic data, etc.
A site where I can input a URL, and it uses's the linked site's sitemap.xml to draw an explorable map (like this one) of all it's pages. Bonus points if you can separate it into "regions" based on topic, time, URL structure, etc.
This is not a complicated idea, but it has lots of fun design avenues to explore. Here's one very simple example.
A plan that sync to your actual waking hours. Starts when you wake up, not at arbitary hour marks.
rotime claims to be a version of this. See also: Superlocal
Use accelerometers and motion sensors to get highly accurate positional data, and use it to record the paths a person has taken.
The hard part will probably be combining this with GPS to place it on an actual map.
I'm sure there are many reasons that this is a bad idea, but I want a service with the vibes of Tinyletter and pricing that isn't a monthly subscription.
A real scenario: you want to move your newsletter off of Substack. It has a few hundred subscribers but you don't want to pay $9/month for emails you only send out 4 times a year (sorry, we both know it's true).
App Bloat leaderboard – A database of popular (pro/con)sumer apps and how high their memory usage and bundle size is.
Where Should I Host? – A comparision of compute and storage providers for the rest of us.
App reviews – Would be cool to make high-quality reviews of apps and games, that aren't just some Techcrunch article.
From aidan: instead of a wall of text it says this is a PC-3PAds-Arbitration site
I like this idea because anybody can start doing this today. Make up a privacy license and add it to your site/product. If it's a good one, other people might see it and borrow the name and terms.
Use their code, shove it into a
Tauri wrapper, set it up to save files to my disk. I will pay up to a $100
one-time fee for this, if done well.
This has been done, but you can do it yourself if you want (hint: Tauri was a bad idea, you'll need Electron)
You know what AI agent I would actually use?
Look through my bookmarks -> navigate 1-2 layers through thread and links -> embed everything for me to search
Minh Nhat Nguyen
This is one place embeddings search might actually be useful because tweet-sized posts mean chunking is already done for you, and the dataset is not large enough to create too many confabulations.
Don't ask for specifics, either you see the vision or you don't.
First step, run continuous backups of the browser history (since Chrome only stores 90 days worth). Then, do fun things like data viz, link re-surfacing, LLM-enabled recommendations, and whatever else you can think of.
Gather the best reading lists, warning posts, useful snippets, field reports, etc. that currently live in the dark forest of Github's gists pages.
Then organise them by niche and host them on a microsite.
While I won't be a user, I would rather people try new ideas than try to make another version of the same old thing.
This very page is a step towards what I want it to be, but it's missing a bunch of stuff.
The Workshop and the Storefront
It felt like people were offering us display cabinets when we were looking for a sturdy workbench. Building a Design System Workbench
This makes more sense in AR, but I would like one in 2D too. Let me arrange folders on a canvas, with icons that open the folder in Finder when clicked on. Let me place things over or under other things, let me stack trash in a corner.
Preserving the web is still an unsolved problem. You can't trust Internet Archive. Most pages are not Designed to Last