BlockWall

BlockWall is a micro-SaaS: renting visual blocks with Web3.

👉 X (Twitter): @BlockWallApp

This post introduces the project and dives into the way I built it: a vibecoding sprint with ChatGPT as a copilot, strict constraints, and fast iteration.

I. Why Web3?

Web3 is becoming a bigger part of our lives. The technology takes time to be adopted, but its advantages are clear:

Speed & reach. Instant, global payments with low friction. No card forms, no chargebacks, no platform lock-in.
Proof & portability. Each rental is tied to a wallet; we store the on-chain signature hash in the database for auditability.
Composability. Wallet identity and on-chain flows let me add rewards, airdrops, or resale later without re-architecting.

Also, I’m a blockchain and cryptocurrency adoptant!

II. The Idea

Plenty of mosaic/ad-wall experiments exist.

The common issues: slow/centralized checkout, limited media (GIF/video is an afterthought), opaque pricing and weak analytics, little love for the dev/creator experience.

BlockWall leans on simple UX (50 blocks, varied sizes: square/portrait/landscape), transparent pricing (size × duration, + small premium for video storage) and a modern stack so I can ship fast and fix things quickly.

Short backstory: I had this idea 10 years ago while living in Bordeaux, France. Back then, a coworker told me it was a stupid idea. Was it inspired by “The Million Dollar Homepage”? Maybe!

III. Vibecoding

What is Vibecoding? It’s a minimal planning, short feedback loops, constraints as fuel, and a copilot mindset with LLM (Large Language Model). Not cowboy coding — but deliberately optimizing for momentum.

The rules I set for my copilot buddy:

One-page MVP. No navigation rabbit holes.
Strict timebox. Ship fast, polish later.
Copilot-first. Use ChatGPT for scaffolding, API hints, and refactors.
Constraints over features. If a choice slows me (SSR/Web3 bundling, for example), pick the path that keeps me moving.

My small rituals in this project:

Micro-sprints (60–90 min): each ends with something visible (layout, upload, rent flow).
“Ask the compiler” quickly: if a build/config fights me for more than 15 minutes, reduce scope or change approach.
Write once, reuse: snippets become composables (e.g., useWallet, useRent).
Make errors loud: good console logs, small toasts, and a “pending → done/failed” state always visible.

How ChatGPT was useful:

Layout generation: went from a generic grid to a deterministic “masonry-like” layout with square/portrait/landscape blocks.
RLS & SQL: drafted tables, views, and no-overlap constraints with proper exceptions (status=‘done’).
S3 flow: presigned PUTs, public reads, cache headers, lifecycle rules; a small downscale step client-side (except GIFs).
Web3 hygiene: moved all Solana client code to .client.ts composables + dynamic imports only inside functions; server-side switched to pure JSON-RPC to avoid bundler drama.
Prod snags: fixed Vercel ESM quirks (e.g., jayson alias), ditched public mainnet RPC for Helius, and cleaned Vite caches.

Thanks to vibecoding & LLMs, it took me ~1 week of coding to get this project up. Without that approach, it would have taken 10× the time!

IV. Tech stack

Frontend

Nuxt 4 and Tailwind.

I’m in love with this combo. Nuxt is amazing and allows you to develop whatever you want, in a meaningful, simple way. Seriously give it a try! Nuxt lets you build meaningfully and simply—seriously, give it a try. Tailwind beats Bootstrap for me, no contest.

Blockchain

I built the dApp on Solana.

It’s robust, scalable, and handles far more requests than Ethereum. Fees are so low it feels criminal.

Server

I used the server that ships with Nuxt (Nitro). There are JSON-RPC calls to the configured RPC (Helius) — no @solana/web3.js on the server. Why Helius? Because the basic public RPC of Solana is so slow, the transaction takes forever or can be reverted.

Database

Supabase is fantastic! You’re on Postgres—but it’s Postgres on steroids: Cron jobs, Views, Constraints, Auth, Storage, Vault, and more.

Others

Storage is on Amazon S3—but any major cloud would work. Deploys are on Vercel, though other providers would do fine.

For better performances: downscale images (not GIFs), lazy images/video, aggressive caching for assets.

V. What was hard (and how I fixed it)

Here is an extensive list of my differents fails and what i’ve done to fix them:

RPC 403 in production (public Solana endpoint): switched to Helius with env-based RPC URLs.
SSR/ESM bundling fights: made web3 strictly client-only; server uses fetch JSON-RPC.
Mobile taps opening the wrong modal: stop propagation on rented blocks, sane z-index, “only fire select if available”.
S3 ACL mismatch: no ACLs; rely on bucket policies + presigned PUT.
Slow media: client downscale (except GIF), Cache-Control: public, max-age=31536000, immutable.

Conclusion

BlockWall is tiny, but it hits a sweet spot: a fun canvas for creators/brands + serious plumbing underneath (payments, storage, RLS). The vibecoding approach let me ship in record time without sacrificing stability. And I really enjoyed creating it.

Try it now