Wading Into the Swamp
10/1/2026
When I got into messing around with server admin in the late 90s the infrastructure was physical and the status I chased focused on the idea of running a dedicated server. An actual box in a datacenter where if the hard drive failed everything else did as well. If you were with it enough to have a backup routine you would still need a human to install and reboot that backup. There was no self-healing, no automated recovery, no resilience - just downtime.
I don't remember what the year was when I first encountered the infrastructure management tool going by the name of Chef. Maybe it was 2012, the year of the Mayan prophecy that didn't really go anywhere (at least that I can tell) or maybe it was a year earlier, maybe later. Either way I do remember that the whole approach to configuration management and infrastructure as code opened a portal in my brain. It was one of those moments when things start to make sense and you realize that managing ephemeral server images as if they were hardware really made no sense. Seems natural now, but back then the oxymoron of infrastructure as code took some getting used to.
Fast forward to 2026 (or so) and AI is in full bloom. Another paradigm shift reconstructing operational mental models across the spectrum. How do you manage the great shift to probabilistic systems? To be sure this is still an evolving shift. There are no simple easy answers or best practice to follow. All we have is smart people and the ability to look at the history of technological change. One of those smart people is Adam Jacob - he was a driving force behind Chef, a founder of Opscode and now is a self-described AI maximalist. His new company Swamp Club is an attempt to bring clarity to the current operational mire. I was re-introduced to his work through the RedMonk Podcast and the conversation immediately caught my attention. I went to swamp-club.com straightaway and created an account.
At that point I had no idea what it was or what I would do with it. And quite honestly I still don't. However my history with Chef served as something of a foundation, a crutch of confidence that let me limp into the swamp. The site is just plain fun to interact with and the community is full of smart people trying new things. After a few months of lurking at the outskirts and a little hiccup in account registration I decided to wade into the swamp with some intention. I decided to build an extension.
At first I had no real sense of an appropriate use case. I looked at the publicly available extensions, but nothing there really clicked for me. I wasn't managing infrastructure in a way that I had an immediate need. So I took my vague sense of utility and bootstrapped Swamp into a monorepo (Turborepo FWIW) I use to manage my personal web site and my daughter's. The repo is pretty simple - a couple of Next.js applications that deploy to Netlify. Everything is static, no database, just markdown files stored in a content directory. On my site I was pulling the latest entries to mturro.com/raw from the write.as API - the service I use to compose those poem-entries. I figured that workflow was as good a place as any to get my feet wet with Swamp. I asked Claude to help write the extension to automatically sync the entries from write.as via a command that I can run manually after I write or incorporate into the deployment process that runs from GitHub to Netlify. The extension was not intended to be helpful to anyone but me; and even in that sense more so to understand the mechanics of Swamp than to actually provide me with utility.
Still, when you're driving a car that lets you go faster than you really know how to drive, it's hard to keep from hitting the gas. At some point either Claude asked me or I asked it about making the extension public. Again, not because I thought it would be valuable to the public or anything more than cruft in the public Swamp Club registry, but because I wanted to understand the mechanics of publishing there. Like most things Claude made it exceedingly easy and now you can see it here.
https://swamp-club.com/extensions/@mturro_29gn/writeas
I can't support it - totally as is, not very useful unless you happen to have a write.as account that you want to pull posts from and repurpose to your own static site - and are doing it all on a Mac (very brittle workflow, but hey, I'm learning).
What I took away from this experiment was the repeatability Swamp introduces to probabilistic workflows - it re-introduces determinism without losing the dynamism of the LLM vibe. Where I am still struggling is with how, where and when to run these workflows. My sense is they are part of the deployment process and can live comfortably in CI/CD world - but there is another nagging sense that while that is true it is also underselling the potential. I feel like there is more to do here, but it's still vague, still forming.
So that's it - my confession. I apologize to the Swamp Club community and the fine folks that build and maintain that registry, but I had an itch. Having now scratched that itch I'm pondering the next itch and wondering how to scratch it. The swamp is calling. Just as Puppet and then Chef did as I tried to wrap my head around the cloud over a decade ago Swamp is helping me adjust to the paradigm shift. The thing that feels familiar to that time is really just the way it dives into an unsettled abstraction (infrastructure as code then, agentic now) and attempts to provide some clarity. The way I feel my mental model morphing is the familiar part. The frustration, the exploration, and the sense that this is meaningful are like old friends.
I can't say that after having built this little toy extension that I have perfect clarity on where things are going, but I had the same experience with Chef back then. Ultimately the most valuable thing this extension experiment did for me was break the seal and put me a little further into the swamp.