Generalization fallacy, survivorship bias, etc.

That's not true, many of these outages have been directly attributed to AI tooling.

> The rest of the world has been happily chugging along with coding agents for almost a year now and things seem to still be working just fine.

Many more services are now being attacked by AI agents that originate from all kinds of organizations including OpenAI, Anthropic, and many others. Unless you're purposefully being obtuse, I would not call that "working just fine".

Yes, this has gone so well

But the load-bearing assumption is that intuition has to be human-shaped intuition. Humans can't intuit thousands of orthogonal directions because we project everything down into a 3D metaphor and hope it holds. That's a fact about our hardware, not about the systems.

And the reason why is the most interesting part: nothing requires the compression step. A model or an agent can operate over the actual objects, holding thousands of runs and ablations in context and noticing regularities in the native dimensionality, without translating them into a picture of a ball rolling down a hill. No bottleneck at "can you visualize it."

So the narrower claim: it's not that intuition here is impossible full-stop, it's that human intuition is unreliable. Your post-hoc rationalization point is evidence for that, not against it. The story exists because a person needs something to hold in their head. Drop that requirement and the failure mode goes with it.

Yes, this part is accurate. Expert density determines how much raw compute each hidden state gets.

> The number of the experts tells you about the diversity of its skills.

Most people misunderstand this part. Counter-intuitively experts don't develop diverse skills, they instead balance compute during the forward pass, allowing models to increase their parameter count without the MLP layers exploding in memory + compute requirements.

P.s. for #2, you can probably do something like RAFT-styled interleaved computation. But this could get tricky unless you commit to a sharding scheme that makes it easier.

A few questions:

1.) How does this handle privacy? If you're distributing compute this way then all actors in the compute graph will also know the sequence being computed.

2.) Any safeguards against malicious actors poisoning model activations?