Jeff Douglas and a co-host define Vibe coding with AgentForce VIBES: prompting AI to build a Salesforce agent from natural language, uploading screenshots and PDFs for the AI to translate into code, or asking it to auto-optimize for higher volume. Writing Apex code manually or handing specs to a developer doesn't count.
To ViVe or not to ViVe? That is the question.
Let's play! Is it ViVe coding or not? Featuring product manager Jeff Douglas. You type, build me a Salesforce agent that handles customer complaints, and AI builds it. That is ViVe coding, yes. A developer writes 300 lines of Apex code from scratch and debugs every error.
That is not ViVe coding. That's what we used to do before ViVe coding.
You upload a screenshot and say, "Make this a working Salesforce app," and AI does it.
That is vibe coding as well. You can definitely do that now with multimodal support in AgentForce VIBES. You can send it an image, a PDF, whatever, and it can translate that into whatever you need to build that application.
A team writes detailed specs and then hands them to a developer to code.
That is not vibe coding. You can take those same specs, you can give it to AgentForce VIBES, and it'll produce the same results.
You tell AgentForce VIBES, "Optimize my agent for 10 times more requests." and it rewrites itself. I would say that's vibe coding also.
It can do a lot of optimization, can give you information on what you would need to optimize things and how you could do it. It may not be something that it can do itself. It may need some other, some other work from other inputs, but that could be some vibe coding. All right, now here's your test. I'm going to give you a prompt and you tell me if it's vibe coding. Create a validation rule for opportunity that prevents it from being closed if the amount is less than $100,000.
