Welcome to the first issue of Tidepool.

We're living in a window where the "Vibe" (the idea) is becoming the "Code" (the product). The gap between "I need a tool that does X" and "the tool exists" just collapsed. This newsletter is about building in that window: some AI news that actually matters for marketers, and exactly what I'm shipping as I go, tool by tool, tidepool by tidepool.

I'm Dustin. Marketing Director by day, builder by night, and most weekends. Zero engineering team. No dev budget. Not much traditional coding experience either. Just Claude Code, a clear problem, and way too many Monsters/Ghosts/Alani etc.

Welcome to Issue #1. Let's get into it.

My entire career. That's how long I’ve worked on teams that run UTMs out of a spreadsheet where half the entries said "LinkedIn" and the other half said "LI." Useless for reporting and frankly a pain in the ass.

I fixed it in 30 minutes.

utmcampaignbuilder.com is a "madlib" style UTM builder — enter your URL, pick from dropdowns for consistency, click copy. Built in Claude Code. Zero budget. Zero procurement. Zero contract review. Zero data passing through a vendor, you'll have to explain to legal someday.

Is it a world-beater? No. But a spreadsheet problem that existed for years is now a solved problem that lives at a URL. That's the point. The bar for "building a tool" used to be months of backlog and engineer time. Now it's an afternoon (or less) and a clear description of the problem.

Try it here → utmcampaignbuilder.com

Someone reached out to me yesterday, trying to pivot into marketing from another field, and asked what they should get certified in. I started rattling off Google AdWords, GA4, HubSpot... and then stopped.

I'm not certified in any of those. I'm a Marketing Director. If I need to pull data from GA4, I pull it and prompt the AI to interpret it. The certification was never the point. The output was.

Melissa Reeve wrote a great piece for MarTech this week about walking into a digital marketing classroom at Colorado State and finding terrified students. Not lazy. Not complacent. Genuinely scared that AI was automating the exact entry-level tasks they were paying tuition to learn. She's right that the fear is valid, and she's right that the jobs aren't disappearing. They're shifting from execution to oversight. Go read it.

But here's where I'd push further: telling new marketers to become "AI auditors" is still playing defense. My advice is more aggressive. If you're a new grad or making a career pivot, build something. Anything. Pick a workflow problem you've seen in any job you've ever had, describe it clearly, and ship a solution in a weekend. Then get it implemented somewhere real. Show a manager. Put it in front of a team. See if people use it.

Because here's what most marketing leaders won't admit: we're all still figuring out what to do with AI. Yes, some are a little further along. But hollistically, it’s early innings for all of us. The person who walks in with a working tool or case study is going to stand out in a way that no amount of coursework can replicate.

Building is half the equation. Getting it implemented and proving time savings is the more important half.

The internship busy work that used to "build character"? That's automatable. Try to automate it. That's the portfolio that matters now. That’s what would matter to me if I were hiring a Marketing Coordinator or Specialist.

→ Read Melissa's piece here

$8 million in ad spend. That's what Perplexity's "Computer" ran autonomously, making 224 "micro-optimizations" during hourly campaign audits. They published a number: $225K in marketing software spend eliminated over a year.

I'm skeptical of the exact figure. What if the AI was sent down the wrong path? What if the ICP assumptions baked in were wrong from the start? That's not a $225K mistake, that's an $8 million one.

But the underlying signal is real: the tools that sit between your data and your decisions are the most vulnerable to replacement. The work that requires human judgment, positioning, and taste? Still safe. For now.

Havas and Broadhead used Claude Code to ship bespoke marketing tools without engineering teams. Broadhead built their entire GEO monitoring platform in a single night. Havas built a Brand Insights AI tool in Replit.

Neither needed a developer.

The moat used to be "we have the budget for engineers." Now the moat is taste and speed. Any marketer who can clearly describe what they want to build can build it. That's a massive unlock and a serious threat to anyone whose value was just owning the tools.

Remember waiting a quarter for dev resources to build a gated ROI calculator?

That's over.

The window we're in right now, where a clear idea and a few hours can produce something real, won't stay this wide forever. The marketers who start building now will have tools, reps, and instincts that are very hard to replicate later.

That's what we're doing here. See you next week.

— Dustin

If you know a marketer who's been waiting for permission to start building, forward this to them. You can subscribe to Tidepool here.