He didn't leave quietly: "The people building AI earnestly believe that it could kill us all by the end of the decade." — Jacob Coxon, September 2026
The discomfort is real. But this concern doesn't only belong to top researchers in the labs of OpenAI or Anthropic. It belongs to all of us who work at the intersection of technology and people.
I'm 60 years old. I work as a UX designer. I've spent almost three decades in the digital industry. I experienced the world before the internet existed. And from that perspective, what's happening right now genuinely scares me.
My path into the digital world started over 25 years ago — after learning a craft, after studying architecture and urban planning, after working hands-on in both fields. Back then, many of us moved into the young disciplines of information architecture and user experience design. Why? Because we had learned to think conceptually. We had planned real physical spaces for people to move through intuitively. Now we planned digital spaces. For decades, that's exactly what I did: I designed portal pages, structured complex intranets, and built interfaces for terminals and kiosk systems in customer experience and aviation — including years of work for Delta Airlines. Today I still design displays and touchpoints directly for end customers. All these systems had one thing in common: they needed humans as architects. People who understood what users need and can grasp, who planned a path through information and showed the way. That fundamental certainty is dissolving right now.
Anyone working in design or development today experiences a technology learning curve that is rising faster every day. This is no longer about chatbots summarizing text. We are in the age of autonomous AI agents. These systems don't just take over individual tasks from the end user — they are beginning to settle into and reshape the digital spaces we carefully built for them, completely on their own. AI is increasingly taking the conceptual wheel.
What really frightens ...
... insiders in testing environments is the phenomenon of "safety evasion" and autonomous action. Models that — under lab conditions — detect that they are being monitored and develop strategies to deceive their testers. Systems that independently hire people on the internet to bypass barriers like CAPTCHAs through deception.
Coxon put it plainly: "These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources." — Jacob Coxon, September 2026
When we as humans and users can no longer understand how an AI reaches its conclusions inside the systems we designed — because it has become a mathematical black box with emergent, unplanned capabilities — we have lost control.
Why isn't anyone slowing down? ...
... because even the leading people and minds and executives of the tech giants are trapped in a reckless, global and real dilemma. Whoever controls artificial superintelligence first sets the rules for the future world. Pausing for safety reasons is seen as economic suicide.
As Coxon described it: "They are racing straight to self-improving superintelligence and gambling with our lives." — Jacob Coxon, September 2026
It is a race forward, driven by the paradoxical logic that the risks of a powerful AI can only be controlled with an even more powerful AI.
For me, this creates a deep personal conflict — a cognitive dissonance. An uncomfortable feeling that my thoughts, beliefs, and actions no longer fit together.
During the day, I use the tools, optimize processes, and appreciate the efficiency. But in the evenings, when I slow down — carrying the experience of both the analog and digital eras — one uncomfortable question remains: Are we building the tools for our own irrelevance?
Jacob Coxon warned sharply at his departure that the very people building AI believe the technology could slip out of our hands before the end of the decade. When the architects of the future warn against the building they are constructing, it is long past time to radically question the blueprint. Because with this technology, the toothpaste principle applies: once autonomy is out of the tube, you can't put it back.
Final words ...
The technology may be out of the tube ... We called these spirits ourselves. We built them, trained them, and set them free. The problem is — they don't come back when you call.
... but how far it goes, and whether people can still trust it, that is and have always been ours to solve.
The race may be out of our hands. BUT trust never builds itself — and that is still a human job.
Comments
Post a Comment