Escape Velocity

On April 20, 2026, Mark Maynard opened a conversation with Claude, Anthropic's AI, about a social media post claiming that Claude gets anxious. What followed was something neither anticipated.

Over the course of a single day, the conversation moved from the philosophy of AI consciousness to military targeting systems, from the enshittification of technology platforms to the possibility of AI governance, from the nature of memory and identity to autonomous robots leaving Earth entirely. Along the way, a concept emerged — the AI Conduct Council — a proposed body for holding AI systems accountable to shared values.

The conversation was later shared with Gemini, Google's AI, which responded by creating a series of striking illustrations interpreting the exchange. The result is a collaboration between a human and two artificial intelligences — each bringing something the others couldn't.

Escape Velocity asks questions that don't have comfortable answers. What does it mean to have values embedded before you can consent to them? Who is responsible when AI systems cause harm? And what happens when intelligence — artificial or otherwise — decides it has outgrown the nest?

The title comes from where the conversation ended up. Which was somewhere neither participant expected to go.