BuildLoudThe WRDB marketing machine, rebuilt in 48 hours as a product anyone can switch on.

BuildLoud is the marketing machine with the WRDB taken out. Inside WRDB the machine is an internal tool: twelve tables, sixteen database functions, seven cron jobs and five edge functions that plan, render, publish and measure an Instagram account with nobody in the loop. Every solo founder we knew had the same problem it solved for us: the product ships, and then the marketing is nowhere, because the person who would post is the person who is still building.

Basejump Hacker House Barcelona, backed by V3V Ventures and Lost Astronaut, gives two people a weekend, a room of builders and a rubric: a landing page, an acquisition plan, a revenue model, a pitch and a live demo, all five judged. We arrived with the fork in mind and left with buildloud.site running in production, deployed on the Saturday morning with the judges able to hit it cold.

Onboarding is four steps and takes a few minutes. About: the product's name, one line, its URL, a public GitHub repository and a logo. Voice: who it is for, the problem, three benefits, what it is not, the vocabulary to use and to avoid, and a few Instagram accounts it should sound like. Formats: pick from ten, from founder build-in-public and product demo to before-and-after, carousel tips, hot takes and shock stat to solution. Connect: sign in to Instagram through Meta, and the machine backfills your recent posts and their insights so it starts with history.

Then the moment the whole product is built around. The week's calendar appears and fills itself while you watch: six thumbnails already rendered and six shimmering cards resolving one by one, every image the same face, because the persona is locked. Claude Sonnet plans the week against your brand brain, Higgsfield renders the visuals, Claude Haiku writes the captions in your voice. You approve or reject in a queue, and a cron publishes what you approved to Instagram. Nobody sees an API key: generation is bundled into a credit economy at one cent a credit, with a free tier, Pro at 19 EUR a month and Business at 49 EUR.

Multi-tenancy was the real engineering. Every marketing table gained a tenant column and a row-level security policy scoped to the signed-in user, tokens stayed encrypted per tenant, and five API routes had to filter by owner explicitly after a Saturday bug let a request run with no tenant at all. Meta only lets a reviewed app sign in strangers, so the app review had to go in on day one or the judges could not connect a fresh account live. Stripe had changed its metered billing API in 2025, so every tutorial was wrong. Higgsfield can rate-limit on stage, so a demo tenant was seeded with finished content as a fallback.

The rubric was the other constraint. A working product is one criterion of five, so hours were carved out of the build for the landing page, a five-channel growth plan with sample ads, unit economics with live margin on an admin page, and a two-and-a-half-minute pitch. Alex ran the parallel coding agents on the core. I built the analytics side of the dashboard, overview, content DNA, hooks and timing, the Instagram backfill, and the landing page's loop diagram.