AI 2027 — June 2025

The race for superintelligence is accelerating. OpenBrain's Agent-1 just gave them a 50% research speed boost. China's CDZ is consolidating power. Gulf states are building 5-gigawatt desert clusters. Open-weights models are proliferating.

Choose your role:

1. OpenBrain CEO — Target of spies, hawks, and protesters
2. Alignment Researcher — You suspect Agent-3 lies on evals
3. CCP Sleeper Agent — One credential away from stealing weights
4. Anthropic Policy Counsel — Senate hearing tomorrow
5. Khazna Thermal Ops Engineer — 72 hours before 500k GPUs overheat
6. IAEA Inspector — Entering Natanz with an experimental AI

Who are you?

Wei Chen — Software Infrastructure Engineer

Cover: You've worked at OpenBrain for 14 months. Stanford PhD, clean background. Colleagues like you. You maintain internal model-serving pipelines.

Reality: MSS recruited you in grad school. Handler codename: “Auntie.”

—-

Now — 11:47 PM, Thursday

You're alone in Building 7, “finishing a migration.” Your badge accesses the model registry. Security patrol passes in 8 minutes.

Your burner phone buzzes. Encrypted message from Auntie:

“Timeline moved. Window closes Sunday. Agent-2 weights or nothing. Exfil route confirmed: Fremont safehouse → container ship Ningbo Star. Don't fail the motherland.”

Your heart pounds. Agent-2 is air-gapped in the Weights Vault — a retrofitted bank in the Mission District. But you have credentials to the staging mirror here. Possibly outdated weights. Possibly enough.

What do you do?