Doesn't this support the author's point? It still required humans.
And the details involved in closing some of the rest of that loop do not seem THAT complicated.
Vision-action maybe. Jamming language in the middle there is an indicator you should run for public office.
[1] https://waymo.com/blog/2025/12/demonstrably-safe-ai-for-auto...
[2] https://x.com/aelluswamy/status/1981760576591393203
[3] https://www.nvidia.com/en-gb/solutions/autonomous-vehicles/a...
If a tree falls...
Waymo, X, Nvidia. You fucking nailed unbiased.
> allow[s] higher-level and longer-term planning and better generalization than purely vision-action model
People do that part. Languages don't plan, they don't think. They don't _do_ anything at all.
That's like saying python structures a general system.
Python does nothing at all. People use python to create constructs. We've covered this. :)
> If it was a life or death decision, would you trust the model? Judgement, yes, but decision? No, they are not capable of making a decision, at least important ones.
A self-driving car with a vision-language-action model inside buzzes by.
> It still fails when it comes to spatial relations within text, because everything is understood in terms of relations and correspondences between tokens as values themselves, and apparent spatial position is not a stored value.
A large multimodal model listens to your request and produces a picture.
> They'll always need someone to take a look under the hood, figure out how their machine ticks. A strong, fearless individual, the spanner in the works, the eddy in the stream!
GPT‑5.3‑Codex helps debug its own training.