It’s less rigid than a command line but much less predictable than either a CLI or a GUI, with the slightest variation in phrasing sometimes producing very different results even on the same model.
Particularly when you throw in agentic capabilities where it can feel like a roll of the dice if the LLM decides to use a special purpose tool or just wings it and spits out its probabilistic best guess.
True the unpredictability sucks right now. We're in a transition stage where the models can understand intent but cannot constrain the output within some executable space reliably.
The bridge would come from layering natural languages interfaces on top of deterministic backends that actually do the tool calling. We already have models fine-tuned to generate JSON schemas. MCP is a good example of this kind of stuff. It discovers tools and how to use them.
Of course, the real bottle neck would be running a model capable of this locally. I can't run any of models actually capable of this on a typical machine. Till then, we're effectively digital serfs.
Particularly when you throw in agentic capabilities where it can feel like a roll of the dice if the LLM decides to use a special purpose tool or just wings it and spits out its probabilistic best guess.