All Stored in the Old Rust Crate

The heat broke last night.

Two weeks of it, 34 degrees some days, no air conditioning, working at the top of a hill where the breeze that usually rattles up from the water just stopped. Labour of love is the right phrase. It was exactly that. In the early hours, when the temperature finally dropped enough to think clearly, I could hear the sound of halyard against mast. I live walking distance from ten thousand boats. Hamble is the centre of the sailing universe in the UK, probably only rivalled by Annapolis in the United States. The vibe is not too dissimilar, actually. Salt and rope and the particular quiet of a marina at 3am.

That sound, halyard on mast, is what the last two weeks have sounded like from the outside. From the inside it has been something else entirely.

Deployed Works, Alpha Candidate

Deployed.works has been in full testing for the past month. Iterating, testing again, iterating again. The kind of cycle that looks like nothing is happening from the outside and feels relentless from the inside.

What I can say now is that we have a product I genuinely believe meets a real global need. Not just for hiring, though it handles the full spectrum from permanent to freelance. The system does something no platform I looked at does: it tells the buyer whether the work they need done can be handled by a tool, an AI agent, an MCP, or a software-as-a-service product, before it ever gets to a human match.

Every platform I evaluated went one of two ways. Either they asked who you need to hire, which is the wrong starting question. Or they offered a directory of tools with reviews that were, on examination, largely meaningless. Deployed Works starts with the outcome. Tell us what success looks like. The system then recommends across the full capability spectrum, human, product, service, before the deterministic matching logic runs at all.

That sequencing is deliberate. The briefing is the product. Everything else follows from getting that right.

Alpha candidate is close. More updates here as it moves.

A Promise to a Little Girl

Four months ago I made a promise I had no business making.

I am not a writer. I do not have children. And yet somehow I agreed to write a storybook for a little girl. I spent approximately thirty seconds wondering what I had committed to, and then I did what I always do.

I wrote a constitution.md. Then a kernel.md. Then I stood up a new Rust engine.

Four hundred commits later I have a governed content machine for a founding storybook. Artwork done. Governance done. It is currently with legal. When it clears, she gets her book.

I do not entirely know why, but somewhere in the middle of those four hundred commits I found myself wishing I had children. That feeling lasted approximately two minutes before I remembered what I was doing and went back to the terminal.

The book will be good. I am certain of that now in a way I was not when I made the promise.

The Unnamed Third

Octeryx OS. Just Hail Mary. And a third system I am going to be deliberately vague about for now.

What I can say is that it interconnects with the other two in ways that only make sense when you see all three together, and that the architecture underneath the connection is what makes the interconnection possible in the first place. More when it is time.

What has become visible through building all three is that using Panamorphix's context infrastructure solely for internal products has an interesting side effect: I think in linear patterns within systems now. Not linearly about systems. Linearly within them, following the graph from node to node as a natural mode of reasoning rather than a deliberate effort.

That is either a skill or a personality change. Possibly both.

Behaviour OS

A few years ago I built a wellbeing platform. I have written about this briefly before. What I have not written about is what came out of it that I did not expect.

We got published. The American Psychological Association and the Journal of Occupational Health Psychology, for literal behaviour change research. That part was planned, or at least hoped for. What was not part of the published study, but was very much part of the platform, was a PhD-led team that helped me develop a sleep intervention system built on a decision tree of considerable complexity.

I built it. It worked. And then I realised that what I had built was not a sleep system. It was a behaviour system that happened to start with sleep. The decision tree was not about one domain. It was about governed packs that could be applied to any wellbeing discipline, any behaviour target, any intervention framework, with the same underlying engine running beneath all of it.

So I rebuilt it properly. Committed, pushed, merged. Architecture docs done. Decks done. And then I mothballed it.

Not because it failed. Because I do not need it yet. In a few months I will. When that moment arrives, the engine will be there, exactly as I left it, ready to be pointed at whatever the problem requires.

The Build, Mothball, Retrieve Cycle

This is how I build. I need something, so I build it. I use it for what it was made for. And then I put it away until the next moment that requires it, which is usually a different moment entirely from the one I originally built it for.

Behaviour OS will connect into Octeryx OS and Just Hail Mary when the time comes. The storybook engine served a promise and will serve whatever comes next from that direction. Deployed Works has been running in various forms for longer than the current version suggests. The old engines never die. They sit in the crate, patient, until they are needed again.

That is also what this site is.

A place to commit the things that matter before they are needed. A record of the thinking, the building, the two minutes of wanting kids, the sound of halyard on mast at 3am when the heat finally breaks.

All stored in the old rust crate.