There's a war happening in developer circles right now, and like most wars, both sides are convinced they're obviously correct while the other lot have lost the plot entirely.
In one corner, you've got the evangelists who speak about vibe coding like it's the Second Coming. They've built an app using nothing but natural language prompts and now they're ready to mass email every recruiter on LinkedIn about the death of traditional programming. In the other corner, you've got the skeptics sharing horror stories—databases deleted despite explicit instructions, security vulnerabilities shipped to production, senior engineers spending more time debugging AI-generated spaghetti than they ever spent writing code themselves.
I've been a developer for over a decade, most of that spent in the glamorous world of NetSuite and ERP implementations. I've seen enough hype cycles to know that the truth usually lives somewhere in the boring middle. So when "vibe coding" became Collins Dictionary's Word of the Year in 2025, I figured I'd actually try the thing before forming an opinion. Revolutionary concept, I know.
My test subject was Claude Code with the NetSuite MCP connector. For the uninitiated, MCP (Model Context Protocol) essentially lets AI assistants connect directly to external systems and actually do things, rather than just pontificating about how things could theoretically be done. It's the difference between having a colleague who gives you advice and a colleague who can actually log into the system and run the report themselves.
The first time I asked Claude to pull franchise profitability data and it just... did it... I sat there for a moment like a Victorian seeing electricity for the first time. Not because the technology was incomprehensible, but because my brain needed to recalibrate what was now possible on a Tuesday afternoon. Reports that would normally involve me writing queries, exporting data, and wrestling with formatting just appeared. Correctly.
Then I tried creating serial numbers directly through the conversation. Having Claude update records and communicate with the POS system while I watched. That's not a party trick for a demo. That's actual business operations happening because I asked for them in plain English.
Now, before this starts sounding like a press release, let me be clear about the friction. I burned through my request limits faster than expected. There were moments of confusion, course corrections, the occasional "that's not what I meant and you know it" exchange that anyone who's worked with AI will recognize. The tools are young. They have rough edges. They will occasionally do something that makes you question whether you should have just done it yourself.
But here's the thing that the doom-and-gloom "vibe coding hangover" pieces miss: for an experienced developer, the learning curve was genuinely shallow. I wasn't fighting the tool. I was directing it. All those years of understanding how systems actually work, knowing what questions to ask, recognizing when output looks wrong—that expertise didn't become obsolete. It became leverage.
The developers I'm worried about aren't the ones experimenting with these tools and hitting limitations. They're the ones who've decided the whole thing is beneath them, or a fad, or something they'll "look into later." Later has a habit of becoming never, and never has a habit of becoming "why does that contractor half my age keep getting the interesting projects?"
So here's my entirely unsolicited suggestion for 2026: stop watching from the sidelines. Not because AI is going to replace you...it probably isn't, at least not in the way the breathless LinkedIn posts suggest. But because the developers who learn to orchestrate these tools effectively are going to operate at a different speed than those who don't. And in a field where productivity has always been the differentiator, that gap is going to show.
I'm not saying you need to rebuild your entire workflow around AI assistants by February. I'm saying pick one thing. One annoying task, one repetitive report, one area where you keep thinking "there has to be a better way." Point Claude at it. See what happens. You might be surprised.
Or it might delete your database. But at least you'll have a good story.
Top comments (0)