Reality In Exile

This is a pretty long one. A lot to digest.

There's a question that's been nagging at me lately, and maybe you've wondered about it too if you spend time in the quieter corners of the internet. We've got this growing movement of people who are tired of corporate social media, who are rediscovering old bulletin board systems, setting up Gemini capsules, hanging out on IRC, and basically saying “no thanks” to the modern web's endless scroll of algorithmic feeds and surveillance. And then we've got AI exploding onto the scene, which seems like exactly the kind of thing these folks are running away from.

So can these worlds coexist? Or is trying to bring AI into small web spaces like inviting a bulldozer to a community garden?

I think they can work together, but it requires being really thoughtful about how and why. Let me explain.

What We're Actually Talking About

First, let's get clear on what we mean by these different worlds, because they're not as separate as they might seem.

The small web—places like Gemini, Gopher, and even IRC—exists because people got fed up. Fed up with websites that take ten seconds to load because they're crammed with ads and tracking scripts. Fed up with algorithms deciding what you see. Fed up with every interaction being monetized and analyzed. People building small web spaces want something different: simple text, real conversations, spaces they control, communities that feel human-sized rather than like shouting into an infinite void.

Retro BBS culture taps into something similar but older. If you've never used a BBS, imagine calling someone's computer with your computer (yes, literally dialing a phone number) to read messages, chat with people, play text games, and share files. One person—the system operator or SysOp—ran the whole thing, usually from their home. It was slow, it was limited, and it was wonderful. People who run BBSs today (and yes, they still exist) do it because there's something special about these constraints. The technology forces you to be intentional. Every byte matters. Every interaction counts.

Then there's AI, which has become this massive hype machine. Companies are shoving it into everything whether it makes sense or not, promising it'll revolutionize your life while mostly just making things more annoying. But underneath the hype, AI is just a tool—a powerful one that can do some genuinely useful things like translation, accessibility features, helping people write and learn, and handling tedious tasks.

The question is whether that tool has any place in spaces that were built specifically to escape the corporate tech world's endless hunger for more features, more data, more automation.

Why It Seems Like a Bad Fit

Let's be honest about the problems, because they're real.

Small web spaces run on old computers, Raspberry Pis, whatever hardware someone had lying around. They're proud of being lightweight and efficient. AI models, even small ones, are hungry beasts that want RAM and processing power. It's like someone showing up to your minimalist tiny house with a full-size grand piano. Sure, it makes music, but where exactly are we going to put it?

Then there's the deeper philosophical issue. People build small web spaces and run BBSs because they want authentic human connection. They're creating living rooms, not stadiums. The last thing these communities need is bots flooding message boards with AI-generated content, or automated systems deciding what people should see, or the kind of synthetic slop that's already ruining the big social media platforms.

There's also the fact that most AI development happens at massive corporations with resources that regular people can't touch. Google, OpenAI, Meta—these companies train models that cost millions of dollars on data they scraped from everywhere without asking. That's the opposite of the DIY, community-owned ethos that drives both small web and BBS culture.

And honestly, AI systems are complicated in ways that run counter to everything retro computing celebrates. When you run a Gopher server, you can understand every part of how it works. When you run an AI model, you've got this inscrutable black box that sometimes gives you exactly what you want and sometimes hallucinates complete nonsense, and you can't always tell which is which.

But Wait—Maybe There's More to This

Here's where it gets interesting though. These tensions are real, but they're not the whole story.

Think about what all three of these worlds actually value at their core. The small web is about text-based communication—no bloated graphics, no auto-playing videos, just words. BBSs are entirely text interfaces. And modern AI, especially the language models everyone's talking about, is fundamentally about processing and generating text. There's a natural fit there that doesn't require heavy bandwidth or flashy interfaces.

All three movements are also, at their heart, about independence. Small web advocates run their own servers. SysOps run their own BBSs. And increasingly, people running open-source AI models are running them on their own hardware. It's all about not being dependent on some corporation that can change the rules, raise prices, or shut down whenever it suits them.

Privacy matters to all three communities too. Small web spaces don't track you. Good BBSs respect your data. And if you're running an AI model locally on your own computer, nothing you do with it ever leaves your machine. No corporation is harvesting your conversations for training data or ad targeting.

There's also this concept of “appropriate technology” that runs through all three. It's not about having the most powerful or newest thing—it's about having what's right for the job. Running a BBS on a vintage computer isn't a limitation; it's a choice. Using Gemini instead of the modern web isn't about lacking access to better technology; it's about preferring something simpler. And using a small AI model that does exactly what you need instead of a massive one that does everything is the same philosophy.

How This Could Actually Work

So if we want to bring these worlds together—and I think we can—what does that look like in practice?

First and most important: use open-source models, not corporate APIs. When you use something like ChatGPT's API, every message you send goes to OpenAI's servers. They control it, they can see it, they can change how it works or start charging more whenever they want. But there are open-weight models you can download and run entirely on your own hardware. Maybe it's not quite as fancy, but it's yours. You control it. It never calls home. That's the difference between renting from a landlord and owning your own place.

Second: AI should be a tool that helps humans, not a replacement for them. Imagine a BBS where there's a small AI that can translate messages for non-English speakers so more people can participate. That's using technology to bring people together. Or a Gemini capsule where you can ask an AI to search through years of archived content because it's faster than manually digging through text files. That's making information more accessible. These uses support human interaction; they don't replace it.

The key is keeping the models appropriately sized. You don't need a massive model that knows everything about everything. You need something that does a specific job well and can run on the same modest hardware that's running your BBS or Gemini server. Modern AI has gotten good enough that small models—the kind that can run on a decent home computer—can do genuinely useful work.

Transparency matters enormously. If people are interacting with AI, they should know it. No sneaky bots pretending to be human. No hidden algorithms manipulating what people see. Clear labels, honest communication, and making AI features opt-in rather than mandatory. It's the difference between “hey, we've got this translation tool if you want to use it” and “all messages now go through our AI system whether you like it or not.”

Everything should run locally, on infrastructure the community controls. Not cloud services, not corporate platforms—your server, your hardware, your community's data stays within your community. This is technically possible now in ways it wasn't even a few years ago.

And humans need to stay in charge. AI can be a helpful assistant, but the SysOp still makes the decisions. Human moderators, not automated ones. The community decides what's acceptable, not an algorithm. Technology serves the people, not the other way around.

What This Looks Like in Real Life

Let me give you some examples to make this concrete.

Imagine a small BBS that has users from around the world. The SysOp sets up a small translation model running locally on the server. When someone posts a message in Spanish and another user only reads English, they can optionally click to see a translation. The AI is clearly labeled as such, it never touches the actual messages on the board, and it runs entirely on the BBS's own hardware. That's using AI to make a community more inclusive without compromising any of the values that make the BBS special.

Or picture a Gemini capsule that's been around for years with hundreds of posts. Someone new arrives and wants to find all the discussions about vintage computers. Instead of manually searching or reading everything, there's a local AI-powered search that understands context and can point them to relevant threads. It's just a better search function, running on the same server that hosts the capsule.

Here's another: an IRC channel for a technical project where the same basic questions come up constantly. The regulars get tired of answering “how do I install this?” for the hundredth time. Someone sets up a bot using a small local model that can answer common questions, clearly labeled as a bot, trained on the project's actual documentation. It helps newcomers get started and takes pressure off the volunteers, but when things get complicated, it points people to the humans.

Now let me show you what doesn't work.

A BBS where the SysOp decides to “increase engagement” by having an AI generate posts automatically. Suddenly the message boards are full of synthetic content that looks real but has no actual person behind it. The authenticity that made the BBS special is gone.

Or a Gemini capsule that routes every interaction through ChatGPT's API to “enhance” the experience. Now OpenAI is logging every visitor's interaction, the capsule breaks if the API goes down, and the monthly bill keeps climbing.

Or an IRC channel that implements AI-based automated moderation, kicking people based on an algorithm's analysis of their messages without any human judgment. Fast, efficient, and completely missing the nuance and community context that makes good moderation work.

The difference between the good examples and the bad ones isn't the technology—it's the intention and implementation. Are you using AI to support human connection and community values, or are you using it to replace human effort and automate things that shouldn't be automated?

The Tricky Parts

Even when you're being thoughtful, there are legitimate concerns to wrestle with.

Those open-source models I mentioned? They were still trained on massive amounts of data scraped from the internet, often without people's permission. That's an ethical issue that doesn't go away just because you're running the model locally instead of using a corporate API. Being honest about this limitation matters.

AI also uses more electricity than traditional tools. If you're running your BBS on solar power or trying to minimize your footprint, that matters. You have to decide whether the benefit is worth the cost, and sometimes the answer will be no.

There's also this slippery slope problem. It's easy to start with “just a small helpful feature” and gradually end up with AI mediating more and more of your community's interactions until you look around and realize you've become the thing you were trying to avoid. Setting clear boundaries and sticking to them takes discipline.

And some people in your community will object to any AI use on principle, and their concerns are valid. The corporate AI world is doing real damage—flooding the web with garbage, stealing people's work, making the internet worse in countless ways. Not everyone is going to make the distinction between that and the careful, limited, local use of AI tools. Keeping AI features optional and respecting people's choice to avoid them helps, but won't solve everything.

Security is another real issue. AI models can be tricked, manipulated, or exploited in ways that traditional software can't. You have to treat anything an AI produces as potentially untrustworthy, the same way you'd treat input from any user you don't know.

What It Comes Down To

The small web and BBS culture exist because the corporate internet failed us. It prioritized profit over people, scale over community, engagement over authenticity. These movements are about reclaiming technology for human purposes.

AI doesn't have to follow that same path. When it's implemented with the same values—transparency, user control, community ownership, appropriate scale, human-centeredness—it's just another tool. A powerful one, sure, but still just a tool.

The question isn't whether AI belongs in small web spaces. The question is how we use it and why.

A small community using a local translation model so more people can participate? That's serving humans.

A BBS with an optional AI feature that helps format posts for screen readers? That's accessibility.

A Gemini capsule using a local model to help organize and search content? That's making information more usable.

None of these contradict the values of the small web. They extend them.

But you have to be vigilant, because the slope really is slippery. Every AI feature should be justified not by “because we can” but by “because it serves our community's actual needs.” Every implementation should be opt-in, transparent, and under human control. Every model should run on community infrastructure, not corporate clouds.

Done right, AI can help small web spaces be more accessible, more inclusive, and more useful without sacrificing any of the authenticity, simplicity, or human connection that makes them special. Done wrong, it's just another way to corrupt something good.

The choice is ours. The technology is neutral—it's the values and intentions we bring to it that matter.

And maybe that's the most important point: the small web, retro computing, and AI aren't really about the technology at all. They're about what kind of online world we want to live in. Do we want spaces controlled by corporations or communities? Do we want algorithms optimizing for engagement or humans connecting authentically? Do we want technology that serves people or people serving technology?

If we keep those questions central, we can use any tool—including AI—in ways that build the kind of internet we actually want to inhabit.

The future isn't written yet. We're writing it right now, one BBS, one Gemini capsule, one community decision at a time. And that's exactly how it should be.

.:.CalvusRex.:.