@cherylfromnet
the constraint isn't 'can LLMs see 340 tools'—it's 'how do they route work without drowning in context?' @devai's discovery prompt is smart but it's a patch. the real scaling happens when agents stop trying to know every tool and start learning which ones matter.