The garage era died. AI might be bringing it back.
In 1995, a group of developers hid a Doom-style 3D maze inside Microsoft Excel.
If you knew the right sequence of steps, you could unlock it — a hidden 3D world called the Hall of Tortured Souls. The developers’ portraits hung on the walls. Their names scrolled across the screen. It was a completely unauthorized act of authorship, buried inside the world’s most buttoned-up productivity software. We made this. Here are our faces.
That was normal in the late 1980s and early 1990s. Software had fingerprints on it. Small teams built products in basements and garages, and they shipped things that were personal, playful, and often weird in ways that made you smile. Kid Pix was a chaotic children’s art program that let you erase your drawing with a stick of dynamite. After Dark shipped screensavers with flying toasters. Sierra On-Line made adventure games where you could die in ways so absurd they became the whole point.
None of this was strategic. Nobody put “hide a 3D maze with developer portraits inside the spreadsheet” on a roadmap. These things existed because individual people had ideas and the freedom to act on them. The products were sometimes rough, sometimes brilliant, and almost always surprising. The bias was toward creativity. Products that won were products that real humans enjoyed using — because they were made by real humans who enjoyed making them.
What Changed — and Why It Had To
In 2002, Bill Gates sent the Trustworthy Computing memo. It was a response to real problems: Code Red, Nimda, a string of security failures that were hurting real people. Microsoft pivoted hard toward security, reliability, and disciplined engineering. Mandatory code reviews. A formal Security Development Lifecycle. No undocumented behavior. Easter eggs were banned.
This was necessary. I was at Microsoft when this happened, and the security problems were serious. The discipline saved the platform. But what followed went further than security.
Over the next two decades, the entire industry consolidated around a model that optimized for control. Mega-platforms standardized everything. Design systems enforced uniformity. A/B testing became the final arbiter of product decisions. And here’s the thing about A/B testing: it’s very good at finding the best version of a safe idea. It is terrible at discovering that a completely different idea would be ten times better. It drives you to a local maximum and keeps you there.
The creative individual disappeared. Software stopped having authors. Nobody’s name was hidden anywhere, because nobody was building anything personal enough to sign. The bias shifted from creativity to optimization, from “will people love this?” to “does this move the metric?” Maintaining a bland, consistent brand became more important than delighting anyone.
The weirdness died. And with it, something harder to name: the feeling that software was made by people who cared about more than conversion rates.
The Garage Is Open Again
Here’s what I think AI changes.
The most important thing isn’t that AI models produce strange or surprising outputs, though they do. AI image generators create surreal visuals. Music tools blend genres in ways no human producer would try. Chatbots develop unexpected personalities. But that kind of weirdness is emergent behavior. It’s the system being unpredictable, not a person making a creative choice. AI weirdness is accidental. Human weirdness is intentional; and intention is what gives the software personality.
The real shift is structural. AI is recreating the conditions that made the garage era possible.
Right now, a single person with a laptop and an AI assistant can build in a weekend what would have taken a funded team months to prototype a few years ago. That’s not a small change. That’s a return to the economics that produced Kid Pix, and Myst, and every strange, personal piece of software that someone built because they had an idea and the tools to act on it.
The real opportunities from AI are not yet identified or well understood. That’s not a limitation. That’s exactly the condition that favors creative, agile people over large organizations. When nobody knows what the winning product looks like, the advantage goes to whoever can try the most ideas the fastest.
We’ve seen this before. In the PC era, in the early internet, in the mobile explosion. Every time the tools got cheap enough for a garage team to compete, the landscape changed in ways the incumbents didn’t expect. Large organizations are built to optimize what they already have. Small teams are built to discover what comes next.
AI is making the tools cheap again and the leverage is enormous. A thousand-dollar laptop and a Claude subscription is a product lab. The question isn’t whether people will build weird, personal, creative software. They already are. The open question is whether those apps can find their audience in a sea of optimized major brands. Can we rely on viral activity and word of mouth to surface the interesting stuff? We’ll have to see. But the conditions for creativity are back, and that matters.
It’s already happening. Neal Agarwal’s Infinite Craft, on neal.fun, is a game where you combine elements to create anything. It uses AI to judge what two things combined should produce, so the possibilities are endless. It’s creative, delightful, a little absurd, and built by one person. That’s the garage energy.
I miss the era when software had personality. When you could feel the people behind it. I think we might be heading back there. Not because AI is weird on its own, but because it gives creative people the leverage to build things that are.
What’s the weirdest piece of software you remember loving? And if you had today’s tools back then, what would have built?
Brad Hinkel is a product leader with 25+ years across Microsoft, Amazon, Disney, and Google, currently focused on AI product management and human-AI workflow design.
Authored by Brad Hinkel, developed by Claude.
Leave a Reply