Hi everyone,
I would like to ask if others have experienced significant output fatigue when working with Grok 4.6 on complex technical tasks, and how you manage model-specific communication styles. I have sometimes replied even "Fine but why you are telling me all this 
"
In deep technical domains, I often find Grok 4.6 outputs spanning multiple pages, which creates high cognitive overhead when trying to extract actionable implementations. By comparison, working with GPT 5.6 Sol model in ChatGPT tends to yield more direct, to-the-point answers for the same types of tasks. That is just example, nothing more nothing less.
I manage agent workflows using of course an AGENTS.md file within an OpenSpec environment—incorporating rules for plain English, domain term expansions, and structured prompts—yet the verbosity remains high.
What specific constraints, negative prompts, or prompt architecture patterns have you found effective in AGENTS.md to force concise, high-signal technical responses without losing technical precision? Any practical examples would be greatly appreciated. Thanks!
edit: sorry for long message, now from myside