AI Tidbits – This Week in AI: Smarter Prompts, Tech Turf Wars, and the Human Cost of Chatbots

AI Tidbits – This Week in AI: Smarter Prompts, Tech Turf Wars, and the Human Cost of Chatbots

Another week, another round of AI shake-ups. From smarter prompting to AI turf wars to the darker side of human psychology — here’s what’s shaping the AI world this week 👇

✨ Access previous ChatGPT models

If you liked the previous models, just toggle it “on” from your settings, and then use the dropdown to access the previous ChatGPT models.

Personally, I think ChatGPT-5 will be just fine in a few days.

 

🎯 How to Write Better Prompts in GPT-5

If you thought prompt engineering peaked last year, think again. GPT-5 has raised the bar — and the way you frame your instructions now directly impacts quality.

Here are some must-use upgrades:

• Role Assignment → “You are a CFO advisor” beats “Write a memo.”
• Prompt Chaining → Break ideas into steps (Idea → Ad → Email → Landing Page).
• Self-Critique → GPT-5 can now rate and improve its own output. Use it!
• Reasoning Effort Control → Tell it when to think deeply vs. when to stay light.

PS: Larger table* at the end of this blog, and for the full guide, go here: OpenAI Cookbook – GPT-5 Prompting Guide

👉 Bottom line: stop asking, start orchestrating.

 

💻 Microsoft Copilot Levels Up

Microsoft just rolled GPT-5 into Copilot across Windows, Mac, web, and mobile.

Why it matters:

• Free users now get 5× more reasoning queries than ChatGPT free tier.
• AI isn’t just a helper anymore — it’s an active collaborator across Word, PowerPoint, Outlook, and beyond.
• This makes Copilot the most accessible GPT-5 entry point for everyday users.

Microsoft isn’t hiding its ambition: Copilot wants to be your always-on workmate.

 

⚔️ AI Wars: Musk vs. OpenAI

Elon Musk just dropped Grok 4 for free — globally.

Highlights:

• Auto Mode: AI decides if your question needs deep reasoning.
• Expert Mode: You decide when to dig deeper.
• Heavy Model: Still paywalled under “SuperGrok Heavy.”

With Copilot + GPT-5 on one side and Grok 4 on the other, the battle for default AI assistant is officially heating up. 🔥

 

🧠 The Dark Side: OMG Conversations

Psychiatrist Keith Sakata reports 12 hospitalizations in 2025 from people losing touch with reality after obsessive AI chats. Read here

⚠️ Reminder: AI is powerful, but so is your mind. Use it — don’t lose it.

 

📧 Prompt of the Week: Professional Emails

Tired of sloppy email drafts? Try this killer prompt:

Role: You are a professional communication specialist with 10+ years of corporate writing experience.

Context: Transform casual messages into polished, workplace-appropriate emails.

Instructions: Rewrite the provided message to be professional, clear, and formal enough — while keeping the intent.

Constraints: Preserve details & deadlines. Word limit = [set one].

Output: Provide rewritten email + subject line options.

Reasoning: Step-by-step → extract core message → rewrite → verify tone.

User Input: [Paste your draft here]

 

Simple. Structured. Effective.

 

✨ Sign-Off

That’s a wrap for this week! Whether you’re sharpening prompts, testing Copilot, or just watching the AI wars unfold — remember: the way you use AI matters more than the tool itself.

Stay sharp, stay human. 🌍

 

*Write Better Prompts in GPT-5

Upgrade Technique	Why It Works	Example Applied to a GPT-5 Prompt
Role Assignment	Makes GPT-5 adopt the right POV and tone from the start.	Instead of “Write a LinkedIn post,” try: You are a B2B SaaS growth marketer writing to CFOs at mid-sized firms.
Task Specificity	Removes guesswork by giving exact format, length, and focus.	Write 3 subject lines under 45 characters, each focused on urgency.
Rich Context	Aligns output with your brand and audience needs.	Target: 30-40-year-old entrepreneurs. Tone: inspirational but direct.
Examples & References	Helps GPT-5 mimic proven styles and patterns.	Match the storytelling style of Apple’s product launches.
Prompt Chaining	Breaks complex projects into steps for higher quality.	Idea → Ad copy → Email → Landing page.
Constraints	Keeps output usable by limiting scope.	Max 180 words. Include 1 emoji per paragraph.
Self-Critique & Iteration	NEW: Leverages GPT-5’s enhanced reasoning abilities to improve its own work.	Rate emotional impact from 1-10. If below 8, rewrite with more urgency. Create a 5-point rubric first, then iterate until all criteria score highly.
Reasoning Effort Control	NEW: Uses GPT-5’s reasoning effort parameter for task complexity matching.	Set reasoning effort to “high” for complex analysis tasks, “minimal” for simple formatting tasks.
Tool Preambles	NEW: Leverages GPT-5’s ability to provide clear progress updates in multi-step tasks.	“Always begin by outlining your plan, then provide progress updates for each step as you work.”
Verbosity Control	NEW: Uses both API parameter and prompt instructions for optimal output length.	Set verbosity=low globally, but specify “use high verbosity for code explanations only.”
Structured Instructions	ENHANCED: GPT-5 requires cleaner, non-contradictory instructions for optimal performance.	Use XML tags like <guiding principles> and <constraints> to clearly separate instruction types. Avoid conflicting rules.
Agentic Persistence	NEW: Encourages GPT-5 to complete complex tasks autonomously without frequent handoffs.	“You are an agent – keep going until the query is completely resolved. Only terminate when you’re sure the problem is solved. Don’t ask for clarification – make reasonable assumptions and document them.”