How To Invest
A financial-education app rebuilt from React to WordPress, with one firm principle: the rules decide the allocation; the AI only explains.
Investing scares those who have never done it — and the market speaks in jargon. That was the first challenge: giving beginners the confidence to take the first step. The second was personal and technical: to use this AI era to build new skills — starting the product with vibecoding (Base44) and then taking it to a full, custom build of my own.
I’ve worked with WordPress for over 15 years, so the choice was obvious: it fit my skills and the project’s needs — especially distribution. We wanted every article and every concept to be indexable and bring compounding organic traffic.
The most important architectural decision was one of trust: the rules decide (archetype + allocation); the LLM only explains. The recommendation logic is deterministic and auditable; Gemini only adds context in plain language, never choosing for you.
From there the constraints wrote themselves: output by asset class only (never named instruments), disclaimers on every result, the AI key always server-side, and GDPR (export/delete data) as a day-one priority.
A native block theme (FSE) and a custom plugin — hti-engine — exposing the REST API: recommendation, profile claim, account and export. Vanilla-JS frontend, with light charts only where they help and PDF generation in PHP. Around the engine there's an ecosystem: a news plugin that aggregates RSS and generates articles, social images and reels; internal and GA4 event tracking; and daily and weekly newsletters wired to Brevo. The content layer — articles and glossary — is editable without code, because it's what feeds SEO.
Live since June 2025 (it started on Base44, today on its own WordPress build). GA4 data (Jan 2025 – Jun 2026) shows around 3,300 active users and 28,369 events, with reach across dozens of countries and content in several languages (PT, EN, ES, FR, DE and more). The questionnaire — the heart of the product — has a bounce rate of just ~3%: those who start, finish. The biggest lesson: choosing the platform was as much a marketing decision as an engineering one.