Each week we'll gather headlines and tips to keep you current with how generative AI affects PR and the world at large. If you have ideas on how to improve the newsletter, let us know!
What You Should Know
Why Vibe Coding Is Finally Useful to Communicators
Last year, “vibe coding” was everywhere. The term — popularized by OpenAI Co-Founder Andrej Karpathy — describes building software by prompting an AI in plain English rather than writing code line by line. Developers experimented, but there was also a notion that vibe coding was overblown. A Bain & Company report last September found that many developers saw little measurable productivity gain from generative AI coding tools. The hype, critics argued, had outrun reality.
They weren’t wrong. Early tools still required technical fluency to fix what they broke. Now, reality has caught up to the hype. Communicators can build the tools they have always wished existed through plain English. You don’t even have to know what Python, JavaScript, Java, or C++ are (they’re some of the most common coding languages), let alone how to write code in them.
The models are getting better at coding, too. Last week, OpenAI released GPT-5.3-Codex, and Anthropic introduced Claude Opus 4.6. Both make meaningful strides in reasoning and code generation. Around them, platforms like Lovable and v0 have matured into something different from last year’s playground experiments. These generate full web applications through natural language. No coding. No understanding of frameworks required.
At the same time, builders like CEO and Co-Founder of OthersideAI Matt Shumer have argued that something fundamental has shifted — that we’re crossing from experimentation into practical utility. For everyone, not just traditional software engineers.
If you can clearly describe what you want, you can start building it. That could mean:
- A microsite for a campaign launch, live in hours instead of weeks
- An internal app trained on your brand voice to review drafts before they go out
- A style-guide checker that flags banned phrases automatically
- A custom media-list organizer
- A dashboard that summarizes coverage and sentiment
This doesn’t mean every experiment will work on the first try. It doesn’t mean governance and security concerns disappear. But the technical barrier is no longer the gating factor it once was.
And the window may not stay this simple for long.
Karpathy himself has begun talking about “agentic engineering” — systems that plan, test, and iterate the code they’ve developed with minimal human direction. That future is still taking shape. But the teams that understand how to build with natural language today will be better positioned when those agents mature.
The moment many people were waiting for may finally be here. If there’s a tool you’ve been wishing existed for your team, this is the time to see if you can build it yourself.
Elsewhere …
- LISTEN: The Real Reason AI Isn’t a bubble, According to a Data Center CEO
- Anthropic Releases Opus 4.6 with new ‘Agent Teams’
- OpenAI Releases GPT-5.3-Codex, a Coding Model that Helped Build Itself
- A New Bill in New York Would Require Disclaimers on AI-Generated News Content
- An Ice Dance Duo Skated to AI Music at the Olympics
Tips and Tricks
🛤️ Getting things back on track
What’s happening: Stay in one AI conversation long enough, and things can start to go haywire. Whether it’s too much conflicting information or one output that derails your train, it can be hard to get the conversation back on track. But it doesn’t mean all is lost.
Try this: If you can pinpoint where things went south, go back, hover the mouse over your last prompt before the problem output, and click the pencil icon to edit it. When you rephrase your prompt, it will generate a new thread starting with that new response. However, it will also remove anything that came afterward. If your conversation had 10 prompts and you go back to edit the seventh, you’re going to lose the last three.
In ChatGPT and Claude, think of this approach like one of those Choose Your Own Adventure books, because you can toggle back and forth between your initial thread that went wrong and your new one using the arrows on the prompt you changed. In other tools, like Gemini, you can edit the prompt, but you don’t get a new path. Whatever you changed is now gone.
Or this: If you want to keep threads separate, you can “branch” a conversation in ChatGPT by hovering over the last quality response, clicking the ellipsis for “More actions,” then “Branch in new chat.” This way, it starts a new conversation but copies all the context up to that point. This is the way to go if your first conversation is still workable, but you want options, or if you change your mind about the direction you want to go.
Quote of the Week
“The skills that make you valuable aren't the ones AI can replicate — like the ability to reason about problems, understand underlying principles, and make intentional design decisions. If you only learn to use specific tools or platforms, you're vulnerable to becoming obsolete every time something new emerges. But if you understand why we design systems the way we do, and how to evaluate whether an automation fits your needs, you become the person making decisions about how and when to deploy these tools rather than just following instructions.”
— Elijah Rivera, Assistant Professor of Computer Science at Brandeis University, in a Q&A on training students for AI-proof careers
How Was This Newsletter?
