This weekend episode of the AI Daily Brief features host Nathan Lambert exploring Professor Ethan Mollick's latest essay The Shape of the Thing, paired with his earlier October 2023 piece The Shape of the Shadow of the Thing. The discussion examines the transition from AI co-intelligence to agentic systems and the power humans have to shape AI's development.

Mollick's analysis traces AI's evolution from the ChatGPT moment through current agentic capabilities, highlighting exponential improvement curves and organizational disruption. The conversation addresses the feeling of destabilization across markets, politics, and society, while arguing against learned helplessness in favor of active engagement with AI transformation.

Lambert critiques campaigns like Jobloss.ai that highlight AI job displacement without proposing solutions, advocating instead for recognizing human agency in shaping AI outcomes. The episode emphasizes that despite the magnitude of change, individuals and organizations retain significant power to influence how AI develops and impacts society.

From Co-Intelligence to Agentic AI Management

The Shape of the Thing identifies a fundamental shift from ChatGPT-era co-intelligence, where humans prompted AI back and forth, to managing AI agents that can complete hours of human work autonomously in minutes.

This transition represents the second major AI capability leap, following the initial ChatGPT/GPT-4 moment, with workable agentic systems plus reasoners marking the current phase as of December 2024.

The change enables radical organizational experimentation, as AI capability now crosses thresholds that unlock entirely new use cases and change perceptions of what AI can accomplish 'sometimes overnight.'

StrongDM's Revolutionary Software Factory Model

A three-person StrongDM team built a 'software factory' with two radical rules: 'Code must not be written by humans and code must not be reviewed by humans.'

Each human engineer allocates amounts equivalent to their salary on AI tokens, at least $1,000 daily, to power coding and testing agents that provide feedback loops until results satisfy AI standards.

The system converts human-written product roadmaps into shipped software without humans touching or seeing underlying code, with only final product review by humans.

The Shape of the Thing emphasizes that 'the particular details of the software factory matter less than the fact that such radical experimentation into how we work is now not only possible, but likely necessary.'

Exponential Improvement and Recursive Self-Enhancement

AI capabilities show consistent exponential growth across multiple benchmarks, from Mollick's 'Otter on a Plane using Wi-Fi' image test to the meter-long tasks graph measuring autonomous work completion.

Recursive self-improvement (RSI) has moved from theoretical concept to explicit roadmap item, with Anthropic's Dario Amodei noting that 'engineers within Anthropic barely write code themselves anymore.'

OpenAI's latest Codex model was described as 'our first model that was instrumental in creating itself,' while Google DeepMind's Demis Hassabis confirmed all major labs actively work on closing the self-improvement loop.

If RSI succeeds, 'the exponential curves we've been watching would get steeper, with an uncertain endpoint,' though bottlenecks in compute, data, or research difficulty remain possible.

Rolling Disruption and Market Destabilization

The Shape of the Thing describes 'rolling and unpredictable environment for AI advances' where capability threshold crossings unlock radical new use cases and change organizational strategies overnight.

February 2025 exemplified this disruption with the Citrini Research financial crisis prediction, Block's 40% layoffs, and the Pentagon-Anthropic conflict over Claude's government use.

Wall Street experiences the 'SaaS-pocalypse' with disproportionate market impacts following each Anthropic feature announcement, while AI forces itself into political consciousness across party lines.

Semi-Analysis's Dylan Patel compared the current moment to 'being in SF like being in Wuhan right before the pandemic - something is happening, it's going to hit everywhere, but so few people know it.'

Rejecting Learned Helplessness and Embracing Agency

Lambert criticizes the 'feigned helplessness' in AI discourse, exemplified by Jobloss.ai's campaign that tracks AI-driven layoffs without proposing solutions or policy remediation.

The Alliance for Secure AI's ad warns 'entire industries could be gone' in 24 months but offers only vague promises to 'hold CEOs accountable' without specific policy proposals.

Despite technical complexity, nearly 7,000 people joined ClawCamp to build AI agents, demonstrating that 'people are not just going to accept AI happening to them.'

The Shape of the Thing concludes that 'we can see The Shape of the Thing now, but we can still influence the thing itself and what it means for all of us' - the window to shape AI may not last long but exists now.

Policy Innovation and Expanding the Overton Window

Public conflicts like the Pentagon-Anthropic dispute, while unseemly, create space for broader AI policy conversations by bringing debates into public view.

Bernie Sanders' proposed moratorium on AI data centers, though Lambert considers it 'short-sighted and ill-conceived,' elevates the conversation and expands policy discourse.

Andrew Yang's CNBC proposal to 'stop taxing workers and tax AI instead' represents the type of radical policy thinking enabled by current disruption, regardless of practical feasibility.

Lambert argues America's 'long history of people not scared of new ideas' becomes crucial when 'as much is up for grabs as it is right now with AI.'