Here we have domain modelling using the Prose programming language. Prose looks to be equivalent to stuff like PDDL (Planning Domain Definition Language). They might actually be exactly equivalent. Though cant say.
The thing is, here the agent automatically creates and debugs the planning domain. Then during run, it seems to be a partial planner which the guest calls an interpreter. Makes sense.
Only thing I am curious about is whether the execution is speculative or not. From what it looks like it isn’t…which means this is an LLM assisted (LLM run) AI Planner!
(Note that AI planning is not related to LLMs…ai planners are deterministic decision making algoriths)
Here we have domain modelling using the Prose programming language. Prose looks to be equivalent to stuff like PDDL (Planning Domain Definition Language). They might actually be exactly equivalent. Though cant say.
The thing is, here the agent automatically creates and debugs the planning domain. Then during run, it seems to be a partial planner which the guest calls an interpreter. Makes sense.
Only thing I am curious about is whether the execution is speculative or not. From what it looks like it isn’t…which means this is an LLM assisted (LLM run) AI Planner!
(Note that AI planning is not related to LLMs…ai planners are deterministic decision making algoriths)
The next generation of declarative programming perhaps... Interesting times.
I‘m more then skeptical. The question is not, if we could do it. The question if we should do it. We are drunken. Drunken by solutions, by ability.
It‘s fine for showing someone for collecting money. Or to find out which presentation is better.
Generating everything is not the way. This is negating responsibility.
Sorry. I‘m not impressed. I‘m frightened.
The shift from telling software how to do something to defining what success looks like may end up being one of the biggest changes in programming
About damn time