Gonçalo Esteves IDEAS FROM THE POD
WORKING PROTOTYPE / 01

AI Futures Atlas

What can go right. What has gone wrong. What we can test next.
Follow a claim through evidence, opposing arguments, safeguards, and verification.

6 curated cases · 9 lifecycle stages

Curated draft · review invited

Why build this when risk databases already exist?

MIT’s AI Risk Navigator already connects risks, incidents, and governance. MITRE ATLAS documents security techniques; NIST offers risk-management guidance. This prototype explores a complementary layer: benefits and harms, competing causal arguments, proposed requirements, independent tests, and a recorded reason for revising or dismissing an idea.

The starting inspiration is Gonçalo’s reading of the Moonshots discussion with Emad Mostaque. An exact episode quotation and timestamp have not yet been verified. This is an independent prototype; no endorsement or collaboration is implied. Dave Blundin’s video catalogue has not been audited here.

Fiction, analogies, community input, and expert predictions can suggest a scenario. They do not establish that it occurred or how likely it is. Each needs a mechanism, a boundary, evidence, and a way to challenge it.

MIT AI Risk Initiative · MITRE ATLAS · NIST AI RMF