This is a submission for the World's Largest Hackathon Writing Challenge: Building with Bolt.
“You don’t need millions to build something powerful. Just an idea, obsession, and execution.”
Hey DEV fam,
Im Ángel, a digital entrepreneur from Spain who blends street culture, branding, and tech. For this hackathon, I wanted to build something real, useful, and with actual impact. That’s how GetFake.ai was born a tool to detect counterfeit luxury products using computer vision and GPT style reasoning.
❓ The Problem
Fakes are everywhere: sneakers, watches, handbags...
Most people cant tell the difference and many get scammed.
I wanted to create something that could instantly tell what’s real and what’s not.
💡 The Idea: GetFake.ai
A mobile-first AI app that does 3 things:
- Scans any luxury item (photo upload)
- Runs a deep analysis using GPT 4 Vision + Gemini
- Searches the web for clues and combines everything into a verdict
It doesnt rely on a fixed fake-database it reasons on the spot.
Real-time. Multi-layered. Brutal accuracy.
⚒️ The Build
- Frontend & Backend: Bolt.new
- AI models: GPT-4 Vision + Gemini
- Storage & Auth: Supabase
- API Logic: Edge Functions
- Bonus: It even detects subtle fake signs like wrong logo alignment, missing serials, off color stitching, and more.
Biggest challenge?
Teaching the AI that “missing data ≠ fake by default”.
We built a reasoning layer that justifies every decision.
🏁 Why This Post Hits the Criteria
Style & Presentation
This post is structured to be visually clean, with clear sections, real world visuals, and a narrative that flows from the problem to the solution plus fully working demo links, cover image and branded identity. Every element reflects the product's UX mindset.
Clarity
No jargon, no filler. Just real talk about how GetFake.ai was conceived, built, improved and tested all within the 30 day hackathon window. The technical stack is explained clearly, and the AI decisions are justified with logic, not hype.
Originality
Instead of a clone or small tool, GetFake.ai is a vision driven solution to a real-world problem (luxury counterfeits). It’s bold, useful, and built from scratch by one person using bleeding-edge tech like GPT 4 Vision + Gemini. It’s not just a demo it’s a real product in motion.
→ This isn’t just a submission — it’s a launch in disguise.
📸 In Action
- Rolex Datejust 36mm
- Dior B23 sneakers
- Louis Vuitton handbags
⚡ Want to try it?
👉 www.getfakeai.site
🚀 What's Next
We’re going beyond the hackathon:
- Authentication passes (like NFTs for real objects)
- Online store integration (scan before you buy)
- Community-driven trust scores
- B2B for marketplaces
I want GetFake.ai to become the Shazam for counterfeit detection.
🙌 Final Thoughts
This hackathon showed me that with the right tools (and pressure), you can build something incredible in 30 days.
If you’re into AI, vision models, or just tired of seeing people get scammed by knockoffs — let’s talk.
Drop a comment. Share your thoughts. Or just come say hi 🤝
Thanks for reading.
🔗 Links
#ai #gpt4 #hackathon #getfakeai #supabase #replicate #visionmodels #startups
Top comments (9)
This post isn't just a write-up. It's a manifesto.
For everyone who’s ever said:
🚀 Props to Bolt for giving us the spark.
This ain’t my last build it’s just the first of many revolutions.
wlhchallenge #bolt #getfakeai #startups #gpt4 #supabase #visionmodels #ai
Soon i will have tthe app full native and uploaded to appstore finnaly
Love the way you focused on actual explainability with the AI’s verdicts, not just pure accuracy. When I built decision logic for AI, edge cases with missing data were always the biggest pain - how do you handle rare or ambiguous brands?
Thanks a lot! Totally agree, explainability is king, especially when trust is on the line. For rare or ambiguous brands, we run a hybrid fallback: first we use perceptual cues via GPT-4 Vision (like stitch irregularities, material mismatch) and we prompt it to actively reason, then trigger a reverse search to scan online presence. If visibility is low, we assign a ‘confidence range’ and let users opt-in for manual review or crowd validation. Transparency > certainty <3
this is extremely impressive the work it takes to get something actually working and solving a real problem blows my mind you ever feel pressure about putting so much trust in ai when there’s money and reputation on the line
Really appreciate that 🙏 it means a lot.
And yes the pressure is absolutely real. When you’re building something that people might rely on to make financial decisions or protect their rep, you cant just plug in a black box model and hope for the best.
Thats why we dont treat AI as a magical oracle, but as a structured observer:
We force it to reason step by step, justify every insight, and back it up with data or visual evidence.
Even then, we dont let it decide alone we add confidence ranges, optional human review, and crowd signals when things get sketchy.
The goal isnt perfection. Its earning trust one decision at a time.
July 4 – Builders Note 💭
Today I scrolled through the featured projects again... and I couldn’t find mine there.
At first, I ll admit it stung a little. But honestly? I didnt start building my app to chase rankings or praise. I built it because I love this world. I love what we’re creating with AI, the boldness, the weirdness, the breakthroughs.
I’m here to learn from the best, to grow as a dev and a creator, and to surround myself with people who also believe in building the future one insane idea at a time.
I hope someday soon my app gets to sit among the ones I’ve admired on this page. Until then, I ll keep showing up, keep building, and keep falling in love with the process.
Much love to all the rebels, tinkerers, late night coders and pixel warriors out here.
Angel
Hey everyone! 👋
I submitted GetFake.ai to the Bolt Gallery a few days ago, but I haven’t seen it featured yet.
Really proud of what we built for this challenge combining multi-layered AI product authentication with an advanced reasoning engine. Would love to see it highlighted alongside the other amazing projects in the Gallery 💥
If anyone else's project made it into the Gallery already, feel free to drop it here! Let's support each other and make our builds visible 🚀
BuildWithBolt #BoltHackathon #AI #Hackathon #Bolt
🔥 Just dropped this after 30 intense days building GetFake.ai! Would love to hear your honest thoughts or any killer ideas to push it further. What would you add if this was your project?
Let’s make this go wild together. Im replying to everyone 🧠💬
-#devchallenge #wlhchallenge #bolt #ai