AI Policy Wiki
Dashboard

The Shape of the Thing

medium confidence · updated 2026-06-06

Mollick on rolling disruption: exponential AI improvement, radical work experiments (Software Factory), recursive self-improvement, and the single chaotic week of Feb 2026.

"The Shape of the Thing" is a synthesis essay by Ethan Mollick, published on his One Useful Thing newsletter on 2026-03-12. Mollick argues that the broad form of what AI is becoming is now visible, and that three forces are combining to produce what he calls "rolling and unpredictable disruption": exponential capability growth, radical new ways of organizing work, and the explicit pursuit of recursive self-improvement by every major AI lab.

Summary of argument

Mollick frames the essay as his most ambitious attempt to "see the Thing" that AI is becoming. The argument rests on three converging forces. The first is continued exponential improvement in capability. The second is the emergence of experiments that reorganize work around AI rather than around people, illustrated by the "Software Factory." The third is the open pursuit of recursive self-improvement, which Mollick describes as having moved from speculation to an explicit item on the roadmap of every major AI company. He closes with the position that there remains a window to influence the outcome: "The window to shape the Thing may not last long, but it is here now."

Riding the exponential

Mollick argues that multiple diverse benchmarks — GPQA, GDPval, Humanity's Last Exam, and PPBench — show the same exponential improvement curve with "few signs of slowdown." He treats METR time horizons as the clearest single metric, citing consistent exponential gains from GPT-3 through GPT-5. He notes that the "jagged frontier" persists, with AI remaining excellent at some tasks and poor at others, but argues that the frontier advances on all axes simultaneously.

The Software Factory

Mollick describes a "Software Factory" built by StrongDM around two rules: "Code must not be written by humans" and "Code must not be reviewed by humans." In this setup, AI coding agents build software from product roadmaps, AI testing agents build simulated customer environments and provide feedback, and humans never see the code. He reports that each engineer spends amounts equivalent to their salary — roughly $1,000/day — on AI tokens. Willison and Dan Shapiro observed the Factory in operation and reported on it publicly. Mollick argues that the particular details matter less than the fact that "such radical experimentation into how we work is now not only possible, but likely necessary."

The rolling disruption week (Feb 22–27, 2026)

Mollick uses a single week in late February 2026 to illustrate the three forces colliding. On February 22, Citrini Research published the fictional financial scenario about AI-driven economic collapse, which Mollick writes "struck a nerve on Wall Street, leading to major stock market price shifts." On February 26, Block (Jack Dorsey) announced 40% layoffs implying AI causation; Mollick suggests it was likely that "AI was merely used as cover for large-scale layoffs." On February 27, the Anthropic-DoW conflict erupted over Claude's military use restrictions.

Recursive self-improvement on the roadmap

Mollick argues that recursive self-improvement is now openly pursued by the major labs. He quotes Amodei that "engineers within Anthropic barely write code themselves anymore," that the feedback loop is "gathering steam month by month," and that AI may be "only 1–2 years away from a point where the current generation of AI autonomously builds the next." He cites OpenAI's claim that GPT-5.3 Codex "was instrumental in creating itself," and Demis Hassabis's statement that "closing the self-improvement loop is something all the major labs are actively working on." Mollick's own conclusion: "We don't know how far this goes... but I also think we are past the point where recursive self-improvement is science fiction. Instead, it is an explicit item on the roadmap of every major AI company."

Provenance

Published by Ethan Mollick on One Useful Thing, 2026-03-12. The essay advances original synthesis and argument rather than reporting a single news event, and is treated as a position piece: its forecasts (such as recursive self-improvement being "1–2 years away") are attributed to the speakers Mollick quotes rather than treated as established facts.

Relationships

  • related: Recursive Self-Improvement (RSI) — a direct source for the argument that recursive self-improvement is "explicitly on the roadmap" of all major labs.
  • related: AI Race Dynamics — the Software Factory as an illustration of competitive pressure driving radical organizational change.
  • related: AI Labor Disruption — the rolling disruption week as an early real-world example of AI-driven market reactions.
  • related: Agentic AI — the Software Factory as a far-reaching implementation of agentic AI in the workplace.
  • related: The 2028 Global Intelligence Crisis — Mollick situates the Citrini scenario within the broader context of rolling disruption.
  • related: Clawed — the Anthropic-DoW conflict as part of the same February week.