Pendulum: disparity and utopia
Oscillation between two extremes. Unequal realities. Human-in-the-loop means slow. AI agents talking to AI agents means unparalleled speed.
End of 2026 reality: Enterprise software vs Omarchy
At work engineering teams are still in "slo-mo", already utilizing AI, but QA, code reviews and deployment are still very much human driven. Pressure building up around moving faster and faster. "Slo-mo" is for a good reason: accountability and also responsibility remains on the human side.
Outside, it is now possible to file a crash report by the built-in AI agent, which ultimately gets analyzed by another AI agent on the receiving end.
DHH's latest motto, "We will fix everything!" is very encouraging, positive, hopeful and engaging. With that level of enthusiasm, it is believable.
Omarchy holds the promise: AI agents built in, you just need to ask. It feels like a utopia, an end goal, achieved.
No time to (die) investigate
The recent Hugging Face incident flags a serious case: analyzing the results of what took an army of AI agents 5-6 days, takes weeks for a team of analysts to reconstruct and understand. This is the speed we are not prepared to accommodate.
Sounds very much like we crossed that famous singularity line: events are happening at a much faster rate that humans can possibly process.
The solution?
Use AI agents to process the results.
The pendulum
On the one hand we've got a controlled, slow-evolving, understandable artifact - that teams of engineers can understand and govern.
On the other hand we've got AI agents woking with AI agents where we don't even see (care to see?) the code they are writing.
...and... we are hoping for the best.
Seemingly, the software domain is solved now.
No one needs to write code any more, the best, universal, most human programming language is English.
Very well done! Well played.
Until an incident happens. We can't put the blame on software.
What do you think?
