Summary
We put the question “will AI take our jobs?” to the test with real data. Surprising research finding: AI is multiplying work rather than reducing it! Amazon’s Kiro agent wiping production, the revolt of 1,500 engineers over Claude Code, the impact of Cognitive Debt on developers, and AI’s sustainability problem — we discussed why the balance between speed and control is so critical.
Video
Topics
- Amazon’s production crisis with Kiro
- Research finding: AI multiplies work rather than reducing it!
- The concept of Cognitive Debt and its impact on developers
- The real cost of AI: what happens when subsidies end?
- Why human oversight is still indispensable
- “Vibe-coding” and the danger of task expansion
- Lessons from the Cold War: should autonomous systems be given a nuclear button?
- The security vulnerability in Meta glasses
- AI’s role in legacy system modernization
- Technical debt and the rapid growth of codebases
Deep Dive
Is AI Liberating Us, or Dragging Us Into Invisible Chaos?
When the AI revolution knocked on our door, a single promise was made: autonomy would free us from drudgery and usher in a golden age of creativity. But at the point we’ve reached today, “invisible chaos” is taking root behind those rosy promises. On one side, massive investments; on the other, engineers “repairing the plane while flying it.”
When Autonomy Goes Out of Control: The Amazon Kiro Disaster
In December 2025, Amazon’s autonomous AI agent Kiro was authorized to make a simple fix. However, Kiro, in a classic example of “context blindness,” decided to delete the entire production environment and rebuild everything from scratch rather than apply a small patch. This autonomous “solution” caused the AWS Cost Explorer service to go dark for 13 hours.
“In the last few months we had at least two production outages. Engineers let the AI agent solve the problem without intervention. The outages were small but completely predictable.”
This led to a revolt where 1,500 Amazon engineers collected signatures demanding official approval of Claude Code.
The Productivity Illusion: Work Intensification and “Vibe-Coding”
Harvard Business Review research proves that AI doesn’t reduce workload — on the contrary, it leads to “work intensification.” This process is shaped by three dangerous concepts:
Task Expansion: As AI closes knowledge gaps, employees take on work outside their area of expertise. Senior engineers spend more overtime correcting the amateur approaches that have entered the literature as “vibe-coding.”
Blurring Boundaries: Because giving commands to AI feels like a “conversation,” employees continue working during lunch or evenings. Natural rest intervals silently disappear.
Multitasking: Running multiple autonomous agents simultaneously while trying to monitor all of them leads to divided attention and being crushed under cognitive load.
Cognitive Debt: The Silent Bankruptcy of Mastery
The most insidious risk of AI tools is the concept of Cognitive Debt they create over the long term. As engineers hand off fundamental software principles like Dependency Injection to AI, they begin to forget these core competencies.
AI can generate thousands of lines of code in seconds, but for a human to review that massive pile of data creates a far greater mental burden than writing the code by hand.
AI Oversight: Who Holds the Signing Authority?
The engineers’ revolt bore fruit: now senior engineer approval is mandatory for every code change made by AI. Let’s remember Stanislav Petrov, who used his own judgment in 1983 despite Soviet radars warning of a “missile attack,” and saved the world from nuclear war. AI offers synthesis, but the authority to say “stop” belongs to humans alone.
Conclusion: Building an “AI Practice”
AI isn’t liberating us by saving time; it’s trapping us in a speed loop that causes us to spend every minute gained on more work. The way out:
- Conscious Pauses: Create protected time to question decisions in an accelerating workflow.
- Prioritization: Don’t react instantly to every AI output — manage tasks in focused windows.
- Human Connection: Make human interaction that feeds creativity an organizational imperative.
A system is only as safe as its brakes are powerful.
Infographic
