I started building Skysquare on July 4th, 2025. Not for any particular reason other than that, with the holiday, I had given myself the day to do something unrelated to “work”—to do something for the fun of it.
Skysquare was not originally conceived as a business opportunity. It was an experiment: could something that lived only in my imagination actually be turned into reality? The original idea was not terribly different from what Skysquare is today: a companion tab living in my browser that showed me all of the posts on Bluesky about whatever page I happened to be reading.
I came up with the idea because I had grown increasingly tired of scrolling through so much garbage on Bluesky to find anything useful. My own social-media consumption habits had been shaped over many years on Twitter, where I followed a carefully curated collection of writers, publishers, researchers, experts, and other interesting voices whose commentary I relied on for better understanding politics, international affairs, cultural criticism, and whatever else happened to be occupying my attention.
Not that Twitter isn’t full of garbage too—and much more of it now than there used to be—but the core of the information graph I had built for myself there had not fully migrated to Bluesky.
I have always used social media primarily for information discovery: to find articles and ideas that informed people are sharing and discussing within their various interpretive communities; to see what people worth listening to are reading, and what they're thinking about it.
Bluesky, being relatively new and having far fewer users than Twitter, did not yet offer the same depth. Yes, a substantial proportion of its users seem to be writers, academics, journalists, researchers, and other interesting people, but it was not then—and, frankly, still is not yet—a one-to-one replacement for the information environment I had built for myself on Twitter. Even today, when I log into one of my old Twitter accounts, I can often find interesting material more easily there than I can on Bluesky. Whatever else one thinks of the platform, Twitter’s machinery for the particular kind of information discovery I value still works substantively better for me.
It is difficult for me to say precisely why, since I do not know the inner workings of either platform's recommendation systems. But part of it seems related to Bluesky’s early philosophical commitment to giving users more control over what feeds and algorithms they use, rather than organizing the entire experience around a single, heavily optimized recommendation engine from above. That openness has obvious virtues. It also means that discovery can sometimes require more work.
And because I mostly used social media for discovery anyway—and because I kept stumbling across writers I had not even known were on Bluesky whenever I searched for posts about particular articles—I began wondering whether there might be another way to encounter the network altogether.
So I set out to build one.
At first, the goal was almost entirely selfish. I wanted a new way to encounter the commentary already happening on Bluesky, using the things I was reading as the discovery vertical rather than the feed itself.
I had begun to recognize that the Bluesky network was working, for the most part: the right kinds of people were showing up, and interesting conversations were happening. But the Bluesky UX still left me wanting an easier way to find that commentary and discover those voices, so I set out to experiment with an interface that might make the network more useful to me as an information consumer.
One of the magical things about that first prototype was that the highlighting feature worked almost immediately. Suddenly, I could see the passages from the article that people were quoting on Bluesky highlighted directly on the page in front of me. The first time that appeared on my screen, I was delighted. Holy shit, it works! How cool is that?
I had imagined an experience, described it to a machine, and hours later I was using it myself. I fell in love almost immediately with the user experience I had created.
There is something genuinely extraordinary about watching an idea that previously existed only in your imagination come alive in the world. It is one of the most magical experiences I have had with technology, and one of the reasons I remain profoundly bullish about the agentic world now unfolding around us. I suspect that, in the not-too-distant future, this ability—to imagine some small thing, whether software, a physical object, a workflow, or something else entirely, and then simply describe it in ordinary language to an agentic development system and begin making it—will become an ordinary part of intellectual life for an enormous number of people.
I listen every day as people understandably moan and groan about the downsides of agentic systems. I am sympathetic; I went through many of the same emotions when generative AI first began disrupting my own profession. But having now experienced the other side of the equation—having seen what becomes possible once these tools are placed in the hands of someone with ideas but without conventional software-development skills—I often find myself thinking: if they only knew what was possible.
That conviction would eventually become one of the reasons I created ContextQB, where I have begun trying to organize and share some of what I learned over the following year about working with agentic development systems. As Skysquare became more complicated, I found myself constantly inventing new methods for preserving context, structuring projects, handing work from one session to another, and generally keeping increasingly elaborate systems intelligible.
But I am getting ahead of myself.
At this point in the story, I knew almost none of that. I had produced a working prototype in an afternoon, was quite pleased with myself after watching the thing I had imagined suddenly come alive, and—buoyed by what seemed like rather convincing evidence—allowed my mind to wander, saying something to myself like:
Self, you and I both know this content-marketing thing is on its last legs. If you can make a website in a couple of minutes with these tools, the marketplace you intend to address with marketing services is only going to keep shrinking. You know where this is headed. This chapter is coming to a close.
And now that you've seen how easy it is to turn an idea into a working application... dare we pivot into building applications as a service? Or, shit, why not try making our own applications and selling them directly to the customer?
I expect other entrepreneurial types will recognize the impulse immediately, if perhaps not the particular mixture of arrogance and naïveté with which I indulged it. It is the impulse of innovation and progress, exquisitely cultivated by the American cultural logic and, in my own case, reinforced by family inheritance. Both my father and grandfather made careers as small-business owners, so there is something almost literally in my blood that inclines me to believe this kind of thing is possible—to believe that, given enough effort, creativity, persistence, and sheer force of will, I can probably figure out how to build whatever I have decided ought to exist.
And so, predictably, I began doing Pollyanna's math in my head:
What’s the most this could cost? How long could it really take? I built the prototype in an afternoon. Surely, if I gave it a few weeks—maybe a couple of months, tops—I could turn the thing into something real, put it out into the world, and see what happens. By golly!
To be fair, around the same time, I had been noodling on a separate idea for turning one of my own internal batch processes into an application. In fact, the weekend I first started experimenting with Skysquare, I was deep in the middle of a client web-development project and using a small system I had built that integrated Google Gemini to read images and automatically write their metadata into WordPress.
It was exactly the kind of tool that had begun to reshape my sense of what software could do for me. It saved dozens of hours of manual labor per project, and the results were probably better than anything I could have reasonably paid someone else to produce at that scale—or, for that matter, done myself by hand.
So after building a relatively elegant interface for Skysquare in an afternoon, I convinced myself that I could probably turn that work-related system into a proper product too: perhaps a WordPress plugin I could sell.
But before I did that, I figured I needed to prove something to myself first: could I actually build an application that other people would use?
So, emboldened by a weekend of prototyping success, I made a decision. I would wrap up the client project in front of me and build Skysquare in my spare time. And in the worst case, even if Skysquare itself went nowhere, I would come out the other side with the skills and knowledge I needed to turn that WordPress tool into a real product.
It seemed, at the time, like an exceedingly sensible bet.