Further convergence to code

So this is a further convergence to code.

Efficiency, accuracy are still the core issues in agent world.

And code is still the best tool for that.

LLM provided a layer to transform natural language to code and then execute it

We are still building programs, which still need enough details, domain logic and data.

LLM based paradigm enables

More integration through a common standard. More tools are connected to the ecosystem.
Built in information extraction from text, images, video etc.
Built in world knowledge on how to do things.
Built in code execution environment
Automatic error correction and words understanding. Not like rigid syntax.

So we are still building things, just in a different way. We can build more advanced things, and build it faster, through the power of LLM.