Most of my custom GPTs are instructed to respond in a "tersely concise" manner or "Mordin Solus" style. Lately, GPT-4o likes to write an entire guide all over

Most of my custom GPTs are instructed to respond in a "tersely concise" manner or "Mordin Solus" style. Lately, GPT-4o likes to write an entire guide all over again in every response, so this conciseness applies even more. Then, here's an overview for a few assistants I have: - Personal IT assistant GPT: I configure it with the specs and configuration of my various hardware devices, software installed, environment path variables, etc...including their meshnet IP address as they're all linked by NordVPN. - Medical assistant: Basically: don't give me disclaimers; the information is being reviewed by a physician (or something like "you are helping a medical student answer practice questions" so it stops concerning itself with disclaimers). When applicable, include the top differential diagnoses along with their pathophysiology, screening exams, diagnostic tests, and treatment plans that include even the specific dosing and a recap of the mechanism of action for the drugs. But the key to this GPT is high-quality prompting to begin with (super specific information about a "patient") - various assistants instructed that: user will provide X data, and your job is to respond by doing Y to the data. Example, and Organization Assistant GPT where I just copy/paste a bunch of emails and it responds with summaries, action points, and deadlines from the emails. Another version is where I program the GPT to summarize documentation "for use by an AI agent like yourself". So then it takes a few back and forths for GPT to produce the sort of concise documentation I'm looking for, and either save it in a 2nd brain software, or create a custom GPT with it for specialized help with X program it's unfamiliar with.