Weekly reflection on human-AI-things collaborations
The shift isn’t from human coders to AI agents. It’s from coding to engineering.
Every, the company behind Lex (which I’m using right now), released four predictions for 2026. The one that stuck with me: the rise of the “agentic engineer.” A couple of months ago, I wrote about an Austrian developer Peter Steinberger, who described this exact reality—directing a team of AI agents rather than writing code himself.
But calling this “directing agents” undersells what’s happening. The agentic engineer doesn’t just orchestrate; they design the environment in which agents can work at their best capacity. They build their own tools, shape workflows, and define constraints. The craft moves upstream: from writing the code to engineering the context.
This connects to another of Every’s predictions: designers building their own tools. That was what triggered me in the Austrian developer’s account was how much of his work involved creating bespoke tooling for his own process. This is what happens when the friction between idea and prototype disappears—designers no longer depend on a coder to test their thinking. They build, they learn, they iterate. Two principles at work: Understanding by doing, and Eating your own dogfood.
And here’s where it gets interesting: if professionals are building their own tools, will end users follow? Is every product becoming less an interface to data and more a platform for making bespoke tools? Does a new layer emerge where users shape their own outcomes?
I think this points toward co-design—but a richer version than we usually mean. Not just designer and user collaborating, but multiple layers of expertise are woven into the design process. The economist who maps value flows in a community. The ethicist who flags downstream effects. Stakeholders who traditionally appear only in the research phase are becoming part of the building itself.
In this framing, agents don’t just execute—they can represent these roles, simulate options, play out consequences before anything hits the real world. Agents as citizens. Not tools we use, but participants in how we design.
Within Cities of Things we have been looking into human-AI teams for a masterclass on designing these new teams, where this engineering was a key element. Building successful human-AI teams should focus on the relationships among the different human and non-human team members, not on task performance.
This triggered thought is part of my weekly newsletter:
Weeknotes 371 – The real shift is not from human coders to AI agents—it’s from coding to engineering the environment where agents are co-designers. And other news on AI companion devices and robots at CES.
https://www.iskandr.nl/systemic-co-design-with-agentic-engineers
Image: Interpretation by Midjourney