1. Fewer and fewer people can make money via their labor. We will have a small minority of super-winners who magnify their work output 100x with AI, and most people will not be able to contribute, thus will not be able to make money.

2. More and more gains go to those who either own the best AI models (Anthropic, OpenAI, Google), own the infrastructure needed to serve those models (energy infrastructure, datacenters). Everyone else gets access to the benefits of AI only as much as the aforementioned groups cannot better utilize those models.

So far AI models are hard to exploit as a commodity (AI companies benefit more from selling inference than using the inference themselves), but as AI models get better, an AI lab will extract more value from keeping their best models as an internal tool, and sell the outputs of those models. We are seeing the start of that with Anthropic using Claude to make discoveries in bio.

3. AI models become better at capital allocation than humans, so the job of financial planner, hedge fund manager, corporate lawyer, etc. go to swarms of agents, which only the richest will be able to afford. Every loophole, every exploit, every micro-strategy compiled into inscrutable macro-strategy will be utilized by those who can pay for millions of tokens of inference in order to turn their $30 billion into $300 billion. The tactics which humans didn't have the patience, mental energy, or tactical brilliance to employ become just another tool in an infinite toolbox the AI can exploit.

It seems like all these factors spell a perfect storm to an absurdly amplified K-shaped economy, where a very very small minority of entrenched wealth dominates all economic activity, and everyone else becomes a minimally necessary human buffer, kept around for either sentimental or convenience reasons.