(Preface: If your first reaction is to dogpile me for using AI to create code for personal use, please hold your nose and read to the end.)
For over 10 years my content pipeline has worked the same way: content arrives via curated inboxes, which I scan them to queue the likely worthwhile reading; at some point in the future I then read it properly; and if it is worthwhile I annotate, tag and publish it as Stuff I Like. Most content also informs the Stuff I Think or Do precisely because I spent that time annotating them - the process of annotation and publishing is the best way I know of embedding knowledge in my mind and stimulating new ideas.
While that process has stood the test of time, the tools I use have changed constantly — delicious & diigo, Pocket & Raindrop, ifttt & Tumblr, Twitter and others have all been swapped out with better alternatives. The weakest link, however, never changes: getting from adding something to my reading queue to actually reading it.
"A bookmark is a promise you make to a version of yourself who never shows up" - The Rise and Fall of the Second Brain, Chris Ng, 6 August 2026.
I found that quote in my own reading queue - I still haven't gotten around to reading the full article yet. In July that would have been impossible, as back then my reading queue was a Raindrop account, so the only thing I could search was the stuff I'd already processed.
Me: "I built something"
But since then I built ReadingQ Manager, my first Obsidian plugin.
It's simple: I share a URL from phone or browser to my daily Obsidian note; the plugin turns it into a task with deadline=tomorrow before reading the URL to identify which of my projects it's relevant to, tagging it accordingly. As a result, it appears in tomorrow's daily note, a central filterable reading queue, and in the to-do list of every project it touches.
It's just a handful of markdown files in a folder on two PCs and a phone. The code probably isn't elegant, but it runs every day, and I can change it tomorrow. So, no big deal ... if you're a developer.
But this is huge to me, because I'm not a developer. I've spent the last 3 decades describing applications for others to build, including websites I don't have a build budget for. Being able to create software, even if it is just for myself, is a game-changer.
Loki: "No you didn't"
Except, of course, I didn't build it - Loki did. Loki's my perpetual Claude Code agent, built as a result of Peter Kaminski's generosity. I call it Loki because it seems pretty powerful, and I don't entirely trust it.
Loki's now well on the way to rebuilding the rest of the content pipeline onto Obsidian and - via my Personal Data Server - out onto the Atmosphere, so that my Hub becomes something I own (rather than log into) while also being part of a 40+m user social universe.
"Begone, satan!"
Before you condemn me for using AI, remember:
this is personal code, a few Markdown files in a directory inside an Obsidian vault on 2 PCs and a phone: I'm not releasing it to the world, I'm not committing to maintain it, and it doesn't matter that the code isn't elegant
this could not exist otherwise: I'm not a developer, and I'm sick and tired of describing systems I'll never see.
Even then, I'm conflicted and still thinking my way through this. Technology isn't neutral: the way AI is being concentrated into a few massive companies is the very opposite of healthy, while the environmental costs are huge. On the other hand, the outpouring of creativity and productivity from letting everyone build software could be literally world-changing.
It's still too early to tell how these costs and benefits line up, but I do know that I can't not know something about this, and the only way of really internalising what this technology allows is to kick the tyres myself. So expect more news from my own personal coalface.
I also know that if I do ever release something, it'll only ever be a proof of concept, designed to show people what the ideas in my head actually look like. If that leads to enough interest to raise funds, those funds will go to actual developers to sort out the slop.
Like more detail? There are some on my wiki, which will one day hopefully end up on my PDS, too.