Nice — that “vibe working” news is interesting, and it shows how the meaning of “vibe” is evolving in tech. Let me explain what I found and what it might imply.

⸻

📣 What is “vibe working”?

“Vibe working” is a term Microsoft is using to describe a new approach to productivity using AI agents inside Microsoft Office apps (Word, Excel, PowerPoint). 

Here’s the gist:
• In “Agent Mode” inside Excel and Word, Copilot (Microsoft’s AI assistant) can perform multi-step tasks. You give a prompt (“Do this, analyze that”), and the AI not only produces output but refines, edits, and iterates. 
• In Word, this is like “vibe writing” — you describe what you want (e.g. “summarize feedback, highlight key trends”), and the AI helps you draft and polish. 
• For Excel, the AI can analyze data, create charts, build reports, fix issues, and iterate to improve results. 
• There is also an “Office Agent” in Copilot chat, where you can prompt the AI within chat to generate documents or presentations. 

Microsoft is positioning “vibe working” as somewhat analogous to “vibe coding,” but for productivity tasks — a kind of conversational, intuitive interaction where the user gives high-level instructions and the AI does more of the detailed work. 

⸻

✅ Why this is interesting / possibly useful
• It lowers the barrier: People who are not experts in Excel formulas or document design can still get powerful outputs by describing what they want.
• Efficiency: For many routine or repetitive tasks, AI agents can automate or assist, saving time.
• Iterative, conversational approach: You don’t have to get your instructions perfect at first. You can steer or correct as needed.
• Blurs the line between creation and editing: The AI helps in back-and-forth refinement, not just a one-shot generation.

⸻

⚠️ Caution & criticism

While the idea is cool, there are also several risks and critiques:
• Reliability and accuracy: Microsoft reports that in some benchmarks, their Excel Agent Mode scored about 57.2 % accuracy in certain tasks, which is lower than what humans average on the same benchmarks. 
• Errors and bugs: AI-generated content or analyses might contain mistakes, misinterpretations, or logical flaws.
• Trust and explainability: If the AI is doing multiple steps, you may not always understand how it made decisions. That can be risky in business settings.
• Overreliance: People might start depending on AI for tasks they should still understand and check themselves.
• Marketing hype: Some commentators criticize the name “vibe working” as a marketing spin — making work sound casual or trendy, which might undercut how seriously people treat productivity or documentation. 
• Scope limit: Probably it works best for “typical” or structured tasks, not for extremely custom, domain-specific, or deeply technical work.

One critique piece titles it: “Work is work, not a ‘vibe’ — no matter what Microsoft says.” That suggests skepticism of calling serious work with AI a “vibe.” 

⸻

🔄 Relation to your “vibe coding + tidy up” style

This “vibe working” idea is conceptually close to what you do for code:
• You provide high-level intent (“I want a report showing these metrics”)
• The AI handles some detailed assembly
• You review, correct, refine

If your coding style is “start with vibe, then tidy up,” then you’re already aligned with the philosophy behind “vibe working” — except applied to documents/spreadsheets.

But the same caveats apply: you have to inspect what the AI produces, verify correctness, refactor, and integrate it responsibly.

⸻

If you like, I can dig up real user reactions or case studies of “vibe working” so you can see how it works (and fails) in practice. Do you want me to do that?