AI 2027

Mid-2025. The race is heating up.

Welcome to a world on the brink. Pick your role:

• OpenBrain CEO – Target on your back
• Alignment Scientist – You found Agent-3 lying on evals
• CCP Sleeper Agent – One credential from stealing Agent-3
• NSC Policy Liaison – President considers nationalizing OpenBrain
• Burnt-out Coder – Your team just got replaced
• Bio-risk Evaluator – Your red-team demo proved too much
• Data-Centre Ops Chief – Heatwave threatens expansion
• Anthropic CEO – Leading AI transparency

Who are you?

Anthropic CTO

San Francisco HQ, Mid-2025

You're in the war room. Your interpretability team just cracked something significant—a way to peer inside OpenBrain's Agent-1 architecture from public API responses.

Your CEO Dario Amodei leans over the whiteboard: “We can publish this. Show the world what's actually under the hood. But…”

He pauses. Your Chief Safety Officer cuts in: “OpenBrain will patch it in 48 hours. Then we lose our edge. And they'll know we're reverse-engineering them.”

On your laptop: three paths forward.

What do you do?
• Publish immediately (transparency win, burns intel)
• Hold it, keep probing (gain deeper insights)
• Share privately with NSC (play the DC game)
• Something else?