I vibe coded a product this week. Here's what it does and why I did it.

Amazon DSP creative rejection is a tax on your time, and the way it plays out in practice can be messy.

Agencies and brands are constantly going back and forth on rejected creatives, and advertisers may only discover a rejection minutes before a campaign is set to go live. There is also no institutional intelligence built into the process. Every rejection is treated in isolation, inside a black box, with no memory of what came before. We have heard from advertisers, and even people at Amazon, that they’ve been genuinely surprised when a creative has been rejected, or discover that something slipped through the crack after a campaign was already running.

I knew solving this would not get prioritized on our product roadmap, and it did not have a natural home in the product yet, even though we are beginning to build creative tooling in Q2. So I spun up a Replit account and over nights and weekends for the past two weeks, vibe coded the Ad Policy Checker.

What it does

Upload your image or video creative and get an instant verdict before you even touch Amazon's review queue. The checker evaluates across all 8 of Amazon's creative acceptance policy categories: brand visibility, image quality, background and border, text and copy, star ratings, animation effects, prohibited content, and creative quality. It also reviews pixel dimensions and asset file size against Amazon's technical specifications. For each one, you get specific findings and a clear fix, not just a pass/fail score.

Video is fully supported. Most compliance tools only handle static images. Under the hood, we compress the video and extract frames at half-second intervals, then pass that full sequence to Anthropic's most capable vision model, for analysis. Rather than evaluating a single still, the model sees the full motion of the ad: transitions, animated text, any effects that would only appear mid-play. Nothing gets missed because it happened between frames.

The system also learns. After every analysis, we ask one question: what did Amazon actually decide? That feedback gets stored, and the model uses it to calibrate future assessments across all 8 policy categories independently. It weights recent feedback more heavily, distinguishes between errors that were too strict versus too lenient, and adjusts accordingly. The more real outcomes people share, the sharper it gets.

Your first analysis requires no signup. After that, drop your name and email for unlimited access, plus batch analysis up to 10 images and 3 videos at once.

Try it here: https://ad-policy-checker.gigico.ai/

Why I built this

Gigi is a 14-person company. Every hour I spend on something is an hour I am not spending on something else. So this needs a real answer.

Two reasons.

First, I think building with AI is no longer optional for anyone leading an AI company. Not understanding it conceptually. Actually building. There is a difference between delegating to AI tools and being proficient with them, and proficiency only comes from reps. Vibe coding a value-creating product without any engineering support is a rep. I do not think I can credibly sell an agentic AI product if I am not personally operating at the edge of what these tools can do.

Second, I wanted to try and build a founder-led GTM engine. I am the sales team at Gigi. I run outbound, manage inbound, and build our channel partnerships. But like every role that AI is now reshaping, founder-led sales is expanding beyond what it used to mean. It used to be: I sell, the team builds, repeat. A tool like this would have required product and engineering resources not long ago. Today it is squarely inside the GTM function and entirely founder-owned. If this product works the way I think it will, it generates leads, increases awareness of what we are building at Gigi, and scales inbound demand in a way that compounds over time.

Try it and let us know what Amazon actually decided. You will be making it better for everyone who comes after you.

Note: This was originally published as part of our Cherry Picked newsletter. Subscribe here 🍒