Full Width 
Shortcuts 
Sign Up 
Log In 
Articles 
Reading List More from Applied Cartography
Buttondown's core application is a Django app, and a fairly long-lived one at that — it was, until recently, sporting around seven hundred migration files (five hundred of which were in emails, the "main" module of the app). An engineer pointed out that the majority of our five minute backend test suite was spent not even running the tests but just setting up the database and running all of these migrations in parallel. I had been procrastinating squashing migrations for a while; the last time I did so was around two years ago, when I was being careful to the point of agony by using the official squash tooling offered by Django. Django's official squashing mechanism is clever, but tends to fall down when you have cross-module dependencies, and I lost an entire afternoon to trying to massage things into a workable state. This time, I went with a different tactic: just delete the damn things and start over. (This is something that is inconsiderate if you have lots of folks working on the codebase or you're letting folks self-host the codebase; neither of these apply to us.) rm rf **/migrations/* worked well for speeding up the test suite, but it was insufficient for actually handling things in production. For this, I borrowed a snippet from django-zero-migrations (a library around essentially the same concept): from django.core.management import call_command from django.db.migrations.recorder import MigrationRecorder MigrationRecorder.Migration.objects.all().delete() call_command("migrate", fake=True) And voila. No fuss, no downtime. Deployments are faster; CI is much faster; the codebase is 24K lines lighter. There was no second shoe. If you were like me 24 hours ago, trying to find some vague permission from a stranger to do this the janky way: consider the permission granted. Just take a snapshot of your database beforehand just in case, and rimraf away.
Humane Inc. started in 2018; it raised around $250M over five years before coming out of stealth mode with an AI pin that people did not like very much, and today they announced their sale (or, to be specific, the sale of their patent library) to HP for $116M. Here is a hype video from July 2022, over a year before they ever announced — let alone released! — a product. I don't think we draw many interesting lessons from Humane. They feel like a relic from a younger, more Juicero-drenched era: even while they were in stealth mode there was an obvious perfume of vaporwave about them, and I think there's nothing inherently wrong with taking big, ambitious, VC-subsidized swings at gnarly problems that don't quite pan out. The two relatively novel things that come to mind are: "Huge amounts of capital" is a good way to finance infrastructure, and a poor way to finance design. [1] Any company that releases a $700 product for consumers and then neuters it with two-weeks' notice for non-existential reasons is, in a meaningful way, evil. At risk of veering into ad hominem, I think any company that releases a music video before it releases a product is obviously doomed — and one thing that a nigh-infinite runway tends to do is justify poor decisions under the auspices of long-termism. ↩︎
What a start to the year! January was a bit more flat-packed than I think Haley and I would have otherwise liked, but all with great things. (Highlights include: first weekend trip with Lucy, signing a lease on an office, and Third South's annual offsite.) A roundup of writing from the past month: A quick note on nominative determinism, and where Buttondown got its name Lots of media reviews: Catch Me If You Can, Book of Clouds, Becoming Trader Joe (!), We Own This City, Bringing up Bebe, Planes, Trains, and Automobiles, A small snippet on truncating timedeltas in the Django ORM Not as much fun writing in January than I had in the past few months; I promise my backlog of topics is longer than ever before.
Guillermo posted this recently: What you name your product matters more than people give it credit. It's your first and most universal UI to the world. Designing a good name requires multi-dimensional thinking and is full of edge cases, much like designing software. I first will give credit where credit is due: I spent the first few years thinking "vercel" was phonetically interchangable with "volcel" and therefore fairly irredeemable as a name, but I've since come around to the name a bit as being (and I do not mean this snarkily or negatively!) generically futuristic, like the name of an amoral corporation in a Philip K. Dick novel. A few folks ask every year where the name for Buttondown came from. The answer is unexciting: Its killer feature was Markdown support, so I was trying to find a useful way to play off of that. "Buttondown" evokes, at least for me, the scent and touch of a well-worn OCBD, and that kind of timeless bourgeois aesthetic was what I was going for with the general branding. It was, in retrospect, a good-but-not-great name with two flaws: It's a common term. Setting Google Alerts (et al) for "buttondown" meant a lot of menswear stuff and not a lot of email stuff. Because it's a common term, the .com was an expensive purchase (see Notes on buttondown.com for more on that). We will probably never change the name. It's hard for me to imagine the ROI on a total rebrand like that ever justifying its own cost, and I have a soft spot for it even after all of these years. But all of this is to say: I don't know of any projects that have failed or succeeded because of a name. I would just try to avoid any obvious issues, and follow Seth's advice from 2003.
More in technology
Dean Burnett writing for Science Focus: Your Brain Is Hard-Wired to Avoid Exercise. Here's Why So why, even though we’ve evolved to do it, doesn't everyone enjoy exercise? The baffling complexity of the human brain is to blame. Evolving an ability doesn’t
One of the most common questions I’m asked is, “is information architecture still relevant now that we have AI?” Of course, not everyone puts it like that. Instead, they’ll say things like “we won’t need navigation if we have chat” or “AI will organize the website” or “in a world with smart agents, we won’t need UI” or something like that. The gist is the same: Do we need structured information in a world with AI? My unequivocal answer is yes. But given my pivot, you might think this is self-serving. So I’d better explain my reasoning. What Information Does Let’s start here: what is information? A surprisingly tricky question! “Information” is one of those squishy words we use without fully grasping its meaning. Over years, I’ve narrowed it down to a pithy (and hopefully, practical) definition: Information is a means for skillful decision-making. Imagine you reach a fork in the road. A sign points left to Buenas Peras (750 km) and right to Pelotillehue (650 km). If you want to go to Pelotillehue and trust the sign, you go right. The sign gives you what you need to make a choice. That’s information. Note a few things: Information must be understandable. To make a choice, the actor must understand the options. You’ll know this firsthand if you’ve tried to drive in a country where you can’t read the local language. Informing happens in context. The sign only makes sense at that particular junction and is only useful to an actor trying to get to either destination. (And with the means to do so: a pedestrian is unlikely to care about a destination 650 km away.) The decider needn’t be human. Nothing about this definition says the choice must be made by a person: it could be an autonomous vehicle driving to Pelotillehue; the “sign” could be data in its software. Whether human or artificial, the decider needs information. While information may be derived algorithmically, it’s not arbitrary. The distinction on the sign is relevant and understandable to particular drivers on particular journeys. The sign isn’t at some random point in the road, but at the fork. Someone (something?) must decide what should be shown for the right actor to choose at the right time and place. That’s the essence of information architecture. I’ll argue that current AIs can’t yet do this on their own. But before I do, let’s dive a bit deeper into the practice of IA and what it entails. What Information Architecture Does Few people have heard the phrase “information architecture” at all. Of those who have, many misunderstand what it does. At best, they think IA is about setting rigid top-down categorization schemes for information systems. At worse, they think it’s about drawing site maps. While these things are outcomes of information architecture, they’re not what IA is about. I’ve boiled down IA to three basic activities: Organize information. Set the context. Plan for change. Let’s unpack them: Organize information. People only understand things relative to other things they already understand. We make distinctions between things (Buenas Peras/Pelotillehue) and group others (vehicles, cities, roads, signs.) We grok concepts through clustering and contrasting. Set the context. We don’t experience information in a void. Choices only make sense in particular contexts. Moreover, sets of choices create particular contexts. IA isn’t just about organizing information, but defining contexts that influence how we perceive where we are and what’s on offer. Plan for change. The idea that IA traffics in “static” information structures is misguided. There are no static information structures – only structures that change at different paces. IA enables changes to happen without compromising intended meaning. Basically, this means governance. Taken together, these activities differentiate IA from other disciplines. And although AI changes how they’re done (and experienced), it doesn’t make them obsolete. At least not yet. What Artificial Intelligence Does (and Does Not) For non-human systems to independently organize information, set the context, and plan for change, they’ll need capabilities current AI systems don’t provide. That’s not to say they never will. But I don’t see how current architectures get us there anytime soon. Yes, AI vendors are promising that AGI (artificial general intelligence) is imminent. I’m skeptical. The more I work in the space, the more convinced I am that current architectures won’t scale to AGI. It only seems they might because LLMs are so effective at languaging. ChatGPT is vastly more capable, useful, and sophisticated than ELIZA. But at their core, both work by matching patterns in language rather than developing real-world understanding – a prerequisite for the skills needed to architect information: empathy, planning, goal-oriented behavior, learning and adapting, and improvising when needed. Not to mention embodiment, which is essential for true contextual understanding. AFAIK, current AI systems aren’t close to acquiring these abilities. (But there’s much I don’t know; I’m going by my real-world experience.) This isn’t to say these systems aren’t useful. Far from it! I wouldn’t be betting my career on the technology if I didn’t believe these systems have incredible potential. But I also think expectations that they can effectively organize information, set contexts, and manage change in ways that are truly useful for humans (or even other AIs) are wildly optimistic, if not outright magical thinking. At least with the current technology – which of course, might change. Bottom Line: Play for Today If AGI ever arrives in the way many imagine — fully autonomous systems that reason, plan, and adapt like humans — then yes, we might no longer need human information architects. But that’s an if, not a when. We’re certainly not there yet. So I’m keeping my mind open and learning as much as I can. But I also understand that organizations want to deliver better products and experiences now. Current technologies can help. But they won’t do it on their own. For the foreseeable future, they’ll need guidance. That’s why I’m pivoting to architecting structures that allow organizations to use these systems more effectively. Taking a page from Pascal, you can think of it as “Jorge’s wager”: AGI may be imminent, but we may as well act as if it isn’t. When (if) it arrives, we’ll have bigger issues to deal with. For now, we have amazing systems that can take us a long way – with human architects guiding them.
I’m not trying to send any hate at a random MacRumors forum commentor who I don’t know, so please take this more as me explaining why there’s more skepticism about the Vision Pro than other gen-1 Apple products. From the post: BTW, I also
