I think this is not right.
LLMs are hugely suited to code because it's token heavy and the compiler is the RLHF verifier.
The LLMs are still very bad at concepts.
What we see in code does not translate to other, arbitrary domains.
Legal - yes, due to text heavy content with tons of great examples as well.
Everything else ... not really. At least not LLMs as we understand them today.
An LLM by itself is quirky and prone to error, but that's not where the next wave of innovation is happening. Look at harnesses instead. A harness like a coding agent is an incredible productivity multiplier. Coding agents are just the first type of harness. We should expect businesses everywhere to create other kinds of harnesses, thereby transforming every domain where money can be made from increased efficiency. All one has to do is ask a coding agent to build a different kind of harness.
That just means those systems are not yet sufficiently formalized. I think LLMs are pretty bad at math too but with Lean they are churning out results.
This honestly just sounds like you haven’t really explored the capabilities of the current bleeding edge
Like two or three generations of frontier ago (in the long long ago of last year), I had ChatGPT design a PCB, then Claude a case, then I had China manufacture it on demand and turned it into a real product that shipped to real users. Those chinese factories were mostly staffed by machines, I suspect.
I'm not the only one that had that idea, they're advertising these things as features now.
I don't know what's next, but keep in mind that today's LLM's are the worst they'll ever be.