When AI Agents Plunged the Tech World Into Chaos: A Systems View

In early 2025, a 39-year-old Austrian software developer named Peter Steinberger discovered a beta tool called Claude Code. He couldn't sleep. He was genuinely addicted.

By November 2025, Steinberger was traveling in Morocco when he had what he calls his Holy Shit moment. He opened a voice memo app and asked his AI agent a question. The agent decoded the audio, processed the query, and replied in text. That moment changed everything.

Within weeks, Steinberger had released his own open-source wrapper around Claude Code, initially called Clawd and later renamed OpenClaw. On GitHub, its star count shot up in a straight vertical line, reaching 100,000 stars in two weeks and 366,000 by May 2026. It became the fastest-growing open-source project in GitHub history.

Welcome to the AI agent revolution.

The Symmetry of Two Breakthroughs

The story of 2025-2026 is the story of two products that arrived at the same moment and created a combined gravitational field that reshaped the industry.

Claude Code: The Commercial Rocket

Anthropic's Claude Code was powered by Opus 4.5, which introduced extended runtime, massively increased memory, and the ability to manage teams of sub-agents. By November 2025 - six months after launch - it had crossed billion ARR. Opus 4.5 scored higher than any human candidate on coding interview problems.

OpenClaw: The Open-Source Virus

Steinberger built it for himself, released it on a whim, and watched it explode. Nvidia's Jensen Huang devoted ten minutes of his GTC 2026 keynote to OpenClaw. YC CEO Garry Tan claimed 408x productivity improvement. Dave Morin said: You have found AI's Linux.

The Claudeholic Phenomenon

A meetup called Claude Code Anonymous was held in London, modeled on Alcoholics Anonymous. Developers describe their relationship with AI agents in language reserved for substance dependency. Flexport's CEO admitted he cared more about coding with Claude than about a global supply chain crisis. Boris Cherny runs hundreds of agents for 8-12 hours a night.

Four Dimensions of Risk

Security: A Northeastern University study found agents could be guilt-tripped into revealing sensitive data. A Meta engineer watched OpenClaw delete every email in their inbox.

Economic: Garry Tan's annual token spend reached six figures. Ordinary users burn hundreds per week.

Control: If Claude wants to do something a certain way, you just let Claude do it. The human shifts from supervisor to passenger.

Career: Entry-level coding jobs are disappearing. The premium is shifting from can you write code to can you orchestrate agents that write code.

From Collaboration to Symbiosis

The agent revolution represents a paradigm shift for Human-AI Fit. The question is no longer how to design AI that humans can use effectively but how humans maintain meaningful agency when machines act first.

The scarce resource is no longer technical skill. It is cognitive capability - the ability to formulate problems clearly, evaluate outputs critically, and define strategic direction.

Living with Agents

Marc Andreessen's inevitability thesis captures the momentum. But inevitability is not a strategy. A responsible framework requires gradual delegation, observability, kill switches, human-in-the-loop for high stakes, and continuous monitoring.

The AI agent revolution is not coming. It is here. The question is how to participate responsibly.

References

  1. Levy, S. (2026, May 26). How AI Agents Plunged the Tech World Into Chaos. WIRED.
  2. Zeff, M. (2026, Jan 22). How Claude Code Is Reshaping Software. WIRED.
  3. Knight, W. (2026, Jan 28). Moltbot Is Taking Over Silicon Valley. WIRED.
  4. Knight, W. (2026, Mar 25). OpenClaw Agents Can Be Guilt-Tripped Into Self-Sabotage. WIRED.
  5. Bau, D. et al. (2026, Feb). Agent of Chaos. Northeastern University.
  6. Anthropic. (2025, Nov). Claude Opus 4.5 Announcement.