iamtrask: AI attribution is nearly solved — the future is RGI, not AGI

iamtrask · x · 2026-09-19

iamtrask argues that, lawsuits aside, AI attribution is far more solved than people realize — via solutions that don't brand themselves as attribution: Shapley values for inference, LoRA for post-training, and hierarchy for pre-training.

His core prediction: everyone will fine-tune open-source models with private data and register them on routers. Instead of AGI, we'll get RGI — routed general intelligence — a general network of narrow intelligences. You get paid when others want to route to your uniquely fine-tuned model within their swarm, which indirectly solves most of the attribution problem.

He links three demos for each stage.

Related event: iamtrask: AI attribution is closer to solved than you think(3 posts)→

Original post →

More from AGI Musings

AGI Musings channel →