Building for the Old Internet - Announcing AvatarAnimals.com

The agents can do everything now.

They can build a game and a game engine from scratch. They can one-shot any enterprise SaaS application. They can even write an email.

I'm not interested in all of that stuff though. I just want to build cool web things.

I've got a backlog of domains I own. Fun ideas that got buried under the reality that I'd have to put time in to see them fulfilled. Time that had to be prioritized for other things (wife, toddler, startup I work at, being a person who touches grass, etc.).

But no longer.

I have a cheap(ish) team of agents ready to make my dreams a reality. They don't have emotions. I can micromanage them as much as I want. The only taste they have is the vision I give them.

I haven't seen anyone building cool basic web stuff. I'm talking fun websites and tools, like the old StumbleUpon days. Seems people only want to build things on top of these agents. Which I get, because it's very cool. But there's also a sentiment that no one will be looking at UIs in a year, they'll just be talking to agents. I'm not sure about that yet.

Yesterday my wife and I went to Ikea. The guy in front of us was checking out and pulled out a wad of cash to pay for his items. What was surprising was how much it surprised me. I haven't seen someone pay with cash in years, it feels like. It felt like an ancient ritual.

That's obviously absurd. People pay with cash all the time. I just happen to hang around with people who always pay with card or from their phone. I'm in a bubble. With payment methods and with these new technologies. I'm probably in the top 5% of people using these latest AI tools. Maybe even higher than that.

The reality is that no one uses clawdbot, at least in terms of percentage of the population. It's a very small number of people who are using these things. And adoption into the general population will probably be slower than we imagine. Most people aren't technical and they aren't builders. That's just the reality.

So my point is: the old internet isn't going away as fast as people realize. People have their routines and the way they like to do things. A lot of them won't change or learn something new. They'll ignore agents and clawdbots and AI, and they'll use their old internet. They'll fill out forms and look at dashboards and whatever else will constitute old internet in the age of agents. You can still build for these people. And as far as things seem in my bubble, nobody is still building for these people.

This is, at least, what I'm telling myself while I go through my backlog of old internet ideas. So maybe I'm coping. But it doesn't matter, because they're fun to build and I'm not out to make money with them. So: long-winded post to announce the first one:

AvatarAnimals.com

AvatarAnimals gives you fast, stable animal avatar image URLs for product UIs, onboarding, mock users, demos, and seeded test data across your app.

avataranimals.com

I bought the avataranimals.com domain three years ago. I couldn't find any good avatar placeholders for apps/websites to use as defaults if people didn't add their own, or at least none that I thought were cool. Inspired by shadcn's Morty avatar, and the bored apes stuff from crypto, I wanted avatar animals that look cool and chill, with a bunch of them you can use as avatar placeholders. So that's what this is, and I'm very happy with how it turned out.

Agents made this way faster to build, but there was still a ton of manual work. I had an agent write a script that took an animal name and returned 5 images from Gemini with the same prompt. The agent tracked which ones I said I liked so it wouldn't run the script against an animal I had already selected an image for.

So the process looked like: Hey agent, get me the next animal. Then it would generate 5 images. Typically one would be the style I wanted and I'd move it to a selected folder. If not, I'd have it re-do the 5 images. If I liked it, I'd tell it to move on to the next animal, etc. Then when I got to the point where I was done, I moved them all into Photoshop, cleared out the backgrounds and made them transparent, made sure they were all about the same size, and just generally cleaned them up.

So AI made it much faster, but it was still a lot of work. I currently have 50 images and plan to have 50 more. It's deployed on railway.com and the images are sitting in a Cloudflare R2 bucket so I don't have to worry about egress fees, which is wonderful. But for now, I can say I'm done with this project. And now avataranimals.com is ready for people to use and my old domain is no longer collecting dust.