(0->1)-humans

And the era of the Lobster began...

A brief summary and some thoughts on the conversation between Lex Friedman and Mr. Clawfather (Peter Steinberger) The source is here: OpenClaw: The Viral AI Agent that Broke the Internet - Peter Steinberger | Lex Fridman Podcast #491 (https://lexfridman.com/peter-steinberger-transcript)

I just want to leave this here for history. The era of the Lobster has begun.

lobster2

  1. The Era of Agents We are shifting from chatbots to AI agents. The difference is that an agent doesn't just talk; it acts: it accesses the system, writes code, and fixes bugs in real-time.
  2. Self-Modification OpenClaw can modify its own source code. If it lacks a tool for a task, it creates one. This transforms software from a "frozen" product into a living, growing organism.
  3. Security vs. Convenience Giving AI system access is a risk. Peter emphasizes sandboxing (Docker) and human-in-the-loop control, warning that the user's desire for convenience ("just do it for me") is the biggest security loophole.
  4. AI Training in 2026 AI training today consists of three stages: Pre-training (basic knowledge), Mid-training (logic and reasoning), and Post-training (personality and tool usage). The future lies in synthetic data, where AI trains AI.
  5. The Path to AGI AGI won't be a single giant program. It will be a global network of millions of small agents solving tasks together. We aren't waiting for a "spark" of consciousness; we are building it brick by brick.