I thought my request for my vibe-coding app was simple. That triggered a chain of failures that made me wonder if my past six months of vibe coding were worth it. Has this all been a waste of time?

Within a week, I wound up changing my architecture for nine properties with a much better and more resilient setup than I had before. But it nearly broke me in the process.

My initial request for the app seemed so simple. I asked it to add one more blog post to the 80 already on the site. Instead, it revamped the blog so that the blog was just a single landing page for that post. All the other 80 posts disappeared.

The “revert” button didn't do anything. In the process of trying to fix all of this, much of the rest of my site somehow disappeared. I had been working for months on new enhancements, like a partner recommendation database. Everything was gone.

I looked for any backup points on my app, and then on GitHub, which has a copy of all of my code. I'm not technical enough to master GitHub, so I was using it with the Gemini extension in Chrome, constantly asking Gemini to help me find potential backup files. After hours of this, I recovered a lot of the structure of my site, but it was missing crucial chunks of data. And beyond all this wasted time, I feared similar issues would resurface.

I faced the proverbial "trough of disillusionment" (Gartner's term) head on. Why bother with any of this? Was vibe coding a mistake? Why was I spending time arguing in caps lock with a website editor? Shouldn't there be more to life than this?

What doesn't kill you with learning AI makes you smarter. I'm sure I learned something? ("Right?" I defensively asked myself.) 

I started writing this column when all my sites were broken. As I finished it days later, they’re all better than they were before. Here are a few things I've learned as a non-technical AI builder:

1) Failure is all but guaranteed. Things will break. Critical work will disappear. Hallucinations will surface unexpectedly. Accepting this will allow you to prepare for it. 

2) Try to buy AI tools monthly instead of yearly. There are exceptions to the rule, like for corporate projects where you have a long vetting process and need to make commitments. Also, some annual deals are too good to pass up, especially when it's for a tool you've been using for a while. But be mindful of how fast things change. A year ago, Claude Code wasn't generally available, and most off-the-shelf AI coding tools were Tinker Toys compared to what they are now. For any area of AI that matters to you — coding, data analysis, content generation, etc. — there is no way to predict what the best model will be even a month from now.

3) Back up your work. When using a vibe-coding app, ask it how to build a mix of automated and manual backup tools. Spend whatever credits or cycles you need on this early on. I wound up switching my properties to Lovable, and it built me various backup options, including a regular email that it sends me, and a manual trigger I can use in my custom admin hub. Look for other ways to sync with GitHub, back up files to Google Drive, or use whatever system you're comfortable with.

4) Try other tools periodically. I used Lovable in the past but went with a competitor initially. Once my site crashed, I asked Gemini for opinions, but its feature comparisons were outdated. Trying out Lovable again, I was surprised by how much better and easier it was than before. There are some downsides — every option has tradeoffs — but it’s been a major upgrade. I compared it to Claude Code, which most people tend to prefer right now, but Lovable wound up giving me better recommendations for how to improve my site.

While working on this column and rebuilding my web presence, I was at a Japan Society event in Manhattan featuring an interview with Hiroshi Sakurai, chairman of the Dassai sake brewery. He charmingly shared how in Japan, Dassai is known as the brewery that's made more mistakes than anyone else. It has failed the most. That has led to its resilience, and it has a track record of crafting some of the most delicious spirits on the market.

Failure is inevitable with experiential learning, especially with AI apps that are often temperamental and have a habit of going rogue. Preparing and adapting will not just allow you to survive the latest wave of tech innovation; it will make you more relevant and resilient in the process.

David

— David Berkowitz, Chief Community Officer, Marketecture Media

Who: Agency Strategists, Creative Leads, Freelance Consultants

What: Agency strategists are reporting a homogeneity crisis where generative AI tools produce increasingly predictable and bland creative briefs. To combat this, some agencies are building diverse AI agent personas with distinct personalities and using multiple LLMs to force more "sideways" thinking and original insights.

Why it matters: If everyone uses the same prompts on the same models, every brand strategy will sound identical; the competitive edge now lies in how you tweak the AI to be weird.

(Digiday)

Who: Luxury Marketers, Retail CMOs, Brand Experience Teams

What: Fashion leaders are leaning into AI for backend personalization and operations while keeping human-led storytelling front and center. The expectation is that AI will power invisible personalization layers rather than replace human interaction.

Why it matters: AI doesn’t have to be customer-facing to be valuable, especially in premium categories where brand and emotion still win.

(Vogue)

Who: CMOs, Creative Directors, Brand Managers, Enterprise Agencies

What: Adobe is partnering with NVIDIA and WPP to launch agentic AI systems that can plan, create, and activate on-brand content continuously. These agents use NVIDIA's computing power and Adobe’s Firefly models to ensure that the generated creative assets are commercially safe and strictly adhere to a brand’s identity.

Why it matters: Marketers can now move from manual, one-off campaign assets to always-on personalized content engines that maintain brand governance without constant human intervention.

(NVIDIA)

Liking AI Brief? The Refresh is a weekly recap of all the most important advertising news, written by AdTechGod.

Thanks for reading this week’s AI BRIEF by Marketecture. 

Marketecture Media is committed to bringing you the latest news, opinions, and community updates. 

If you are interested in advertising on this newsletter, our podcasts, The Advertising Forum, or in our Slack community, you can contact us here.