Cloud Free, Linux Free, Container Free, Crap Free

Small and simple computing, the 8by3 way

Somewhere along the way the internet got complicated. Not the good kind of complicated, where a clever protocol lets a million unrelated machines find each other and exchange ideas, but the bad kind, where serving a simple web page involves seventeen layers of abstraction, three SaaS subscriptions, a Kubernetes cluster, and a data residency policy you didn't read. Everyone convinced themselves they needed to build like a hyperscaler.

We think it's time to push back a little. So here's what 8by3 stands for, and why we get out of bed. Four small principles, none of them especially radical, all of them increasingly rare.

Cloud Free

Our software runs on computers we own. In fact, computers somebody else used to own; we bought second hand. In a rack, in an office, in Cape Town. It has a power cable, a network cable, and a row of blinking lights on the front that flicker every time you load this page. Right now, one of them is flickering because of you.

We're not against the cloud as a concept. Distributed computing, redundant data centres, fibre running under the ocean, all of it is genuinely amazing. What we're against is the consolidation. A small handful of US-based hyperscalers now run most of the internet's workloads, and the economics of that arrangement quietly squeeze out everyone else: independent hosts, regional ISPs, hobbyists, small businesses, the people who used to make the web feel like a place rather than a product.

When you rent a machine in someone else's data centre, you don't just rent compute. You rent dependency. You rent a pricing model that gets opaquer every quarter. You rent a control panel that suddenly has an "AI features" tab whether you wanted one or not. You rent the privilege of having your data egress charged at rates that would make a roaming SIM card blush.

Owning hardware, by contrast, is a one-off conversation with your accountant and then a long, peaceful relationship with a machine. Used enterprise gear from three or four years ago is absurdly capable. An enterprise grade server cost that six figures new, can be had for the price of a decent laptop second hand, and it will quietly serve millions of requests a month while drawing less power than a kettle. We run a couple of those. The whole estate cost less than a few months of equivalent rentals.

And running hardware is fun. Pulling network cables through walls is fun. Crimping RJ45 connectors badly the first three times and then getting it right is fun. Watching a 10G fibre link come up between my main router and the garden office for the first time is fun. The lights blink, and each blink represents an actual person somewhere in the world reading something we wrote, or buying something from one of our merchants, or planning a community market. That feedback loop, that physicality, is something the cloud quietly took away. We're taking it back.

We're fully on solar. Cape Town has approximately eleven thousand hours of sunshine a year and a national grid that does not, shall we say, share that enthusiasm for uptime. Off-grid compute isn't just an ideological flex here; it's a practical answer to load-shedding. A few panels, a battery, and less compute can keep small services running through a twelve-hour blackout without anyone noticing.

Linux Free

To be fair, we have nothing against Linux. It's done a tremendous amount of good. Its community is enormous and largely well-intentioned, and it powers most of the world's interesting software. We use Linux-targeted tools every day, and many of our dependencies originate in Linux land. No grievance here.

But a healthy ecosystem is one in which there are viable alternatives. Monoculture is fragile in software for the same reasons it's fragile in farming and ecosystems: when everyone runs the same kernel, a single bad upstream commit, a single security flaw, a single corporate acquisition, can ripple through the entire planet's infrastructure at once. Alternatives are insurance.

So we run FreeBSD and OpenBSD. FreeBSD on the servers, OpenBSD where we want the security model turned up to eleven. There are practical reasons for this beyond ideology. ZFS is a first-class citizen on FreeBSD in a way it still isn't on Linux, and ZFS is genuinely transformative once you've used it: snapshots that are essentially free, compression that pays for itself in I/O savings, send/receive that turns backup into a one-line command, integrity checking that catches silent disk corruption before it eats your data. The whole base system is coherent, documented in one place, and updated as a single unit. The handbook is one of the best pieces of technical writing in existence.

There's a softer benefit too. Choosing a less travelled path forces you to actually understand what you're doing. You can't just copy and paste a Stack Overflow answer that assumes systemd; you have to read the rc.conf, understand the jail config, learn what the pkg system does and doesn't manage. The hour or two that costs you upfront is repaid a thousand times in fewer 3 a.m. surprises.

Supporting alternatives helps keep the internet free and open. It also, not incidentally, helps keep the internet functioning. Diversity isn't just nice; it's load-bearing.

Container Free

Well. Kind of.

The orthodox path in 2026 is to stuff every service into a Docker image, hand the images to Kubernetes, hand Kubernetes to a managed control plane, and hand the bill to whoever is unfortunate enough to be running finance. There are organisations for which this makes sense; if you have hundreds of services and dozens of teams shipping changes constantly, container orchestration earns its keep.

Most of us don't have that problem. Most of us have a handful of services that need to start when the machine boots, restart when they crash, and otherwise stay quietly out of the way. For that, FreeBSD jails are almost embarrassingly good. They've been battle-tested since the year 2000. They isolate processes, networks, and filesystems with the same kernel everyone is already running, no overlay filesystems, no daemon-of-daemons, no YAML pyramids. You define a jail, you start it, it runs. You back up a jail by snapshotting its ZFS dataset. You migrate a jail by sending that snapshot over the network. The whole mental model fits in your head, which means you can reason about it at 3 a.m. when something is on fire.

Crap Free

This last one is the most important and the hardest to define. Let's try.

No bloat. The pages on this site are small. They load quickly. They don't ship four hundred kilobytes of JavaScript to display three paragraphs of text. They render on a ten-year-old phone over a 3G connection. They render on a screen reader. They render in a text browser, if for some reason you want to read us in Lynx. That isn't an accident; it's a deliberate choice to respect your time, your battery, and your data cap.

No tracking. We don't run analytics that follow you around the web. We don't fingerprint your browser. We don't load third-party scripts that exist to siphon behavioural data to advertising networks.

No enshittification. The word is Cory Doctorow's, and the pattern he named is everywhere now: platforms that start out useful to their users, become useful to their business customers, and end up useful only to themselves, milking both sides for the last drop of value before collapsing under their own contempt. We're not a platform of that kind, and we've structured the business deliberately so we never have to become one. No outside investment. No growth-at-all-costs board demanding a hockey-stick. A small team, sustainable revenue, products our customers actually pay for because they actually work, and they actually provide more value to their users than they cost. That's it. That's the whole plan.

Why any of this matters

You could read all of the above as nostalgia, or as performative contrarianism, or as the kind of thing a small company says when it can't afford the cloud. Fair enough; we'd push back a little on each charge, but they're at least understandable.

The deeper reason we build this way is that we think the texture of the internet matters. A texture that is replicated in our society, in how people do business and interact. One made of millions of small, independent, oddly-shaped entities, each one run by a person or a small team who cares about it, is a different kind of place from one made of three hyperscaler and and a handful of billionaire controlled companies. The first one has plants growing through the cracks, weird local cultures, people who answer their own email.

The lights just blinked.