The Information Machine

Today’s edition of my newsletter just went out.

Rohan Paul Twitter · Rohan Paul (@rohanpaul_ai) · 2026-06-26

Rohan Paul's newsletter edition covers OpenAI agent-driven office work research, a 2027 IPO timeline, MIT findings on the AI coding productivity gap, and Qwen's new 35B open-weight world model.

Open original ↗

Appears in

Extraction

Topics: ai-agentsopenaiai-economyopen-weight-modelsai-coding-tools

Claims

  • OpenAI research shows AI agents are beginning to handle most execution in office work scenarios.
  • OpenAI is targeting a 2027 IPO as public markets test whether AI companies merit trillion-dollar valuations before proving durable profits.
  • MIT research found AI tools increased code volume 300% but actual output rose only 30%.
  • Qwen released Qwen-AgentWorld, a 35B open-weight world model trained to understand how environments respond to AI agent actions across terminals, browsers, and OS tools.
  • OpenAI RL research found that training on realistic human situations caused safer, more useful behavior to generalize to untrained tasks.

Key quotes

OpenAI is now leaning toward a 2027 IPO because the public market is testing whether AI giants deserve trillion-dollar prices before they prove durable profits.
Code volume surges by 300%, but output increases by only 30%: The AI dividend meets an awkward reality.
AI has pushed the internet's content machine into a new phase, with books, lawsuits, research papers, apps, and songs now being produced at volumes that old review systems were not built to handle.