There are two places from which you can write about technology.
One is the peak of the curve. The first place anyone reaches, and the one that seduces everyone. You know this terrain well by now: there's a technology that has made it possible for anyone to write like a professional, in the same voice, churning out texts, posts and articles at conveyor-belt speed. The peak is where the smoke machine kicks in, where every announcement is revolutionary, where nobody can tell what's going to stick from what's going to evaporate. It's the loudest place to be — and also the easiest. You just repeat what's being said up there.
The other place is the slope of enlightenment, further down the curve. You get there after the disappointment, or after stepping in just deep enough to see the real difficulty of the work. Here the smoke clears, the early evangelists have crashed, the ideas slow down, and they start to find their actual fit. Things stop being so sexy. And that's where you can start metabolising what was said up at the peak, and start building applications and features that actually contribute. Without all the underbrush.
This newsletter is about demystifying the peak — full of noise, but necessary noise — and giving the disillusionment phase the credit it deserves. We're going to walk through the convergence of several technologies that are redefining how humans and machines identify themselves, authorise each other, and operate together.
What's emerging is a new operating system. Human + Machine.
Human/OS.
A short example first
Identification is a much older problem than we tend to assume. The solution we've always used has three roles: an authority issues a document, the holder carries it, and a verifier checks it when needed.
Roman tax collectors did this with sealed tablets. Medieval guilds did it with their charters of trade. Royal courts issued safe-conducts. Renaissance universities granted bachelor's degrees. The pattern has held for two thousand years: issuer, holder, verifier.
The materials change — parchment, sealed paper, plastic with a magnetic stripe, a chip — but the model has survived intact. Not because it's simple, but because it solves a fundamental human problem: how to trust something about someone you don't know directly.
Almost everything we build in software is exactly that. We model human problems and translate into code patterns that were already working. A wallet emulates a wallet. A transfer emulates handing over cash. A digital signature emulates a written one. New technology, at its best, is a more efficient emulation of something we were already doing.
But not always. Sometimes the new thing emulates nothing. Sometimes it creates a model that didn't exist before. And it succeeds anyway.
That tension — between what emulates and what invents, between what a user recognises immediately and what they have to learn — sits at the centre of how a technology matures.
The purpose
This newsletter covers technologies that right now sit at very different points on the curve, and whose convergence will define how humans and machines identify, authorise, and operate together over the next decade.
AI agents are sitting exactly at the peak. Maximum smoke, opinions in every direction, very few productive use cases. We're on the doorstep of the first serious crashes, and what comes through them is what's worth following. This happens every time — it's evolution doing its housekeeping. How it plays out on this round is anyone's guess.
Physical AI — humanoid robotics, embodied AI, world models — is climbing fast toward the peak. Less smoke than agents for now, but with many of the same underlying problems still unsolved.
Digital identity tech — verifiable credentials, SSI, EUDIW — is on the productive slope. The hype phase is well behind it, the standards have stabilised, and real implementations are starting to appear with regulation underneath them, especially in Europe. A decentralised evolution of that family, blockchain-based identity, is now consolidating: Ethereum Attestation Service, ENS and Sign Protocol are already on the productive slope, though other more general components still need to mature.
Post-quantum cryptography is, for me, the most interesting position on the whole map. Real work is being done, real standards have been published, and the arrival of its impact still feels distant to most of the field — which is precisely what makes it worth reading now.
And cutting across all of the above, there's a transversal axis: leading and building human-AI teams. Helping the human side establish a frame for decision-making under incomplete context, getting better at managing uncertainty, sharpening the focus on system orchestration. We'll go deeper into these concepts — and others we'll introduce along the way — in upcoming editions.
What this newsletter will do is observe how these technologies evolve in their respective curves, and pay particular attention to the points where they intersect. Because the real value isn't going to appear inside any one of them in isolation. It's going to appear where they integrate.
That's why Human/OS lives in the valley, not at the peak.
Welcome.
— Javi
Next edition: five observations on how a technology matures — the smoke at the peak, the catalysts, the hammer and the nail, the counter-force, and the cognitive curve of the user.