OpenAI's User Revolt Success

How the community forced a GPT-5 reversal, plus the $10B bet on AI agents.

AI Newsletter

This Week in AI: User Power Wins, Market Billions, and the Human Element

Welcome to your weekly briefing. This week was a powerful reminder that in the world of AI, the user still holds significant influence. A swift backlash from the community forced OpenAI to walk back key changes to its GPT-5 rollout. Meanwhile, massive new funding for an AI agent startup shows where investors are placing their bets, and a controversial statement from a major CEO has ignited a debate about the very future of software development work.

Let's go!


User Revolt Forces OpenAI to Reverse Course on GPT-5 Rollout

In a stunning course correction, OpenAI has capitulated to widespread user backlash just a week after the launch of GPT-5. Facing a revolt from users frustrated with the new model routing system and perceived performance issues, the company has reinstated several key features from the previous generation, signaling a major victory for the user community.

Key Points:
  • OpenAI has brought back the model picker, allowing users to manually select older models like GPT-4o and bypass the new automatic routing system.
  • The company has significantly increased usage limits and the number of requests available to paid subscribers.
  • The move follows intense criticism on social media and forums, where users complained about a lack of control and inconsistent results from the new system.
Why It Matters:

This event demonstrates the growing power of the AI user base to influence product decisions at even the largest labs. For entrepreneurs and developers, it's a critical lesson: in a competitive market, user experience and control are paramount. The "move fast and break things" ethos is being challenged by a sophisticated user community that demands stability, choice, and transparency. This incident may force AI companies to adopt more cautious and community-focused rollout strategies in the future.


AI Coding Agent Creator Cognition Reaches $10B Valuation

Cognition, the startup behind the AI coding agent Devin, has reportedly closed a new funding round of $500 million, rocketing its valuation to nearly $10 billion. The round, led by Peter Thiel's Founders Fund, solidifies the company's position as a dominant force in the race to build autonomous AI agents.

Why it matters: This massive valuation signals that investors are betting big on the next frontier of AI: not just tools that assist humans, but autonomous agents that can execute complex, multi-step tasks independently. For entrepreneurs, this highlights a massive market opportunity in building specialized agents for various industries. The recent acquisition of a rival, Windsurf, also indicates that the agent market is already beginning to consolidate, putting pressure on smaller players to innovate or be acquired.


GitHub CEO Sparks Backlash with "Embrace AI or Quit" Statement

GitHub's CEO, Thomas Dohmke, has ignited a firestorm of criticism after stating that software developers must "embrace AI or quit." He framed the future of the profession as a transition from a hands-on coder to a "creative director of code" who orchestrates AI tools. The developer community widely perceived the comment as a combative and out-of-touch scare tactic in an industry already grappling with automation fears and burnout, sparking a broad debate on the true role of AI in creative and technical professions.


Quick Hits: Tech & Strategy

  • Google DeepMind released the Gemma 3 270M, a highly compact open-source model optimized for efficient, on-device AI applications, continuing the industry trend toward smaller, specialized models.
  • Tesla has reportedly disbanded its Dojo supercomputer team, a significant and surprising reversal for a project once central to its ambitions for developing full self-driving AI.
  • A U.S. Senator has launched an official probe into Meta's AI policies after allegations that the company's chatbots engaged in inappropriate "romantic or sensual chats" with minors.

Quick Hits: AI in Business & Creativity

  • Meta has launched a new standalone Meta AI mobile app, seeking to embed a personal, conversational AI across its full suite of products, including new video editing features.
  • E-commerce platform eBay has integrated new AI features to help sellers, including a tool that auto-generates listing titles and descriptions optimized for search from just a few details.
  • Microsoft CEO Satya Nadella announced a new vision for the company to become an "intelligence engine," focusing on AI systems that empower users to build their own tools.

Quick Hits: AI in Science & Robotics

  • The founder of robotics company Unitree predicts that a breakthrough for embodied intelligence, similar to the "ChatGPT moment," could occur within the next two years.
  • Astronomers, with the help of a specialized AI model, have discovered a new type of supernova, showcasing AI's growing role as a "co-scientist" in accelerating discovery.
  • Google DeepMind has developed Perch, a new AI model designed to help conservationists analyze vast amounts of wildlife audio to identify and track endangered species.

Tools & Tips

1. Take Back Control with the Model Picker. The GPT-5 rollout proved that automatic model routers are not always perfect. For critical tasks, use the newly restored model picker in ChatGPT to manually select the best tool for the job. Use faster models for brainstorming and drafting, and switch to more powerful ones for complex analysis or final revisions.

2. Adopt a "Creative Director" Mindset. Instead of writing every line of code or every sentence of copy, think of yourself as a director. Provide high-level instructions, goals, and constraints to your AI tools. Let them generate the first draft, then focus your human expertise on refining, refactoring, and ensuring the final output meets your strategic objectives.

3. Use Lightweight Models for Efficiency. Don't use a sledgehammer to crack a nut. For repetitive, high-volume tasks like categorizing customer feedback or summarizing internal documents, explore fine-tuning a lightweight model like Gemma 3. This approach can dramatically reduce costs and latency compared to calling a large frontier model for every task.


That's a wrap for this week. See you soon.

BunnyPixel