I still don't know which exact category Maria falls into yet, but her core concept and purpose are clear.
The goal is to build a centralized repository of scripts, tools, and programs. Instead of relying on an LLM to generate code from scratch, Maria uses the LLM strategically to select the best existing option from this repository.
- Cost Reduction: Avoids the high token consumption of continuous code generation.
- Enhanced Security: Running pre-approved, vetted scripts eliminates the risks of LLM hallucinations.
- Error Minimization: Reduces unexpected bugs by using stable, pre-tested tools.
Ultimately, the user experience is seamless: the user simply asks Maria for the tasks they need, and she handles the rest behind the scenes.