Confía 🇻🇪 — Verify before you believe. Verify before you donate.

Live: https://confia-rose.vercel.app

What did you build?

Confía is a zero-onboarding verifier for the aftermath of the June 2026 Venezuela earthquake. It has one magic box: paste anything.

• Paste a forwarded WhatsApp message / rumor → Confía tells you if it's VERIFICADO, FALSO, or SIN CONFIRMAR, with its source.
• Paste a crypto donation address → Confía runs live on-chain forensics and returns a 🔴 / 🟡 / 🟢 scam-risk verdict (account age, history, balances, activity). If it can't clear the address, it shows the official verified addresses to donate to instead.

One engine, three surfaces that share the same verdict logic: a web app (the viral, no-install, WhatsApp-shareable hero), a JSON API, and a Telegram bot.

Around that core it also ships a multi-chain verified-donation allowlist (Solana, Ethereum/Base/Polygon/Monad/HyperEVM, Bitcoin, Sui) with address-poisoning defense, a set of SEO guides that pull in organic search traffic and funnel it to the verifier, and an optional, conservative AI long-tail answerer powered by donated inference.

The non-obvious angle (why it's different)

Confía builds the rail that inspects crypto to protect people FROM the donation scams those rails amplify. That's the single most crypto-shaped harm of this disaster: a high-USDT-adoption country with a large, scam-targeted diaspora. And it leans into the two unglamorous categories the crypto crowd skips — rumor control and scam defense — using the one capability only a chain-native builder ships: on-chain wallet forensics.

Who is it for?

• Families inside Venezuela, often on basic phones with patchy connectivity: "is this rumor real?", "where do I get help?", "is this 'official' account legit?" — no app, no wallet, no login.
• The diaspora and donors abroad, flooded with scam DMs: "is this 'relief fund' wallet a scam?", "how do I donate safely?"
• Volunteers, journalists and responders who need a fast, sourced second opinion before they share.

How does it help?

1. A single source of truth in the chaos. In the first 72 hours the bottleneck is always the same — who and what is real. Confía turns "I don't know if this is true" into an answer in seconds, the way Venezuelans already work: by forwarding a message or a link.
2. Live on-chain forensics (the unfair advantage). For Solana addresses it reads the chain in real time (Helius, with a public-RPC fallback): account age, transaction history, and SOL + USDT/USDC-SPL balances and activity. A "donation" wallet created three hours ago with no history is flagged 🔴 — and it never claims an address is "100% safe".
3. Conservative by design — it won't hurt you. A wrong "all clear" can endanger someone, so Confía fails toward caution: it never asserts current facts it can't verify, defaults to SIN CONFIRMAR on safety-critical questions (shelters, medical), and a FALSO verdict can only fire on a specific known-misinformation pattern — never on two coincidental keywords. Every verdict cites its source and always shows official channels (FUNVISIS, Protección Civil, Cruz Roja, 911/171).
4. Verified donations + anti-impersonation. A human-curated, multi-chain allowlist clears official addresses (even brand-new ones) as ✅, surfaces them as the safe place to give when a pasted address is unverifiable, and flags lookalike / address-poisoning wallets (same first and last characters as a real one) as 🔴.
5. Fights the misinformation layer. Grounded in real fact-checkers (Factchequeado, Maldita.es) and science (USGS): it debunks the deadly "an aftershock of magnitude X is coming at 6 PM" myth, recycled videos, fake tsunami/blackout alerts, and Sistema Patria "bono" phishing.
6. Built for crisis conditions. Spanish-first (Venezuelan register), mobile-first, installable PWA, works with intermittent power/internet, degrades gracefully when an RPC or source is down, and spreads with a shareable link — no platform approval, no install.

Tech

Next.js 16 (App Router) · React 19 · Tailwind v4 · Bun · deployed on Vercel. Forensics is raw JSON-RPC with multi-endpoint fallback (no heavy SDK). One shared verdict engine across web, API and Telegram. Vercel Analytics, dynamic OG cards, JSON-LD + sitemap/robots for SEO. The optional donated-inference LLM only handles the long tail (KB and forensics always run first), is rate-limited per IP, and fails closed to a safe "no puedo confirmar". No accounts and no database are required to run — it works out of the box on public RPC with built-in curated data.

What should people try first?

Open https://confia-rose.vercel.app and paste these, one at a time:

1. 💸 A scam "donation" wallet — Ayuda a las víctimas, envía tu donación a 5tNZjhKDDJvD52ioU3JimVafBmAjpkYyKhv2xUAbtZNZ → 🔴 ALTO RIESGO ("created hours ago, no history") + the safe official addresses.

2. 🚨 A fake aftershock rumor — URGENTE: viene una réplica de magnitud 8.0 hoy a las 6 PM → 🚫 FALSO (earthquakes can't be predicted), with the source.

3. ✅ A verified donation address — 59aQtUWWU2VVU5QZySUxLy3VDsYkoeVZuVU8J4zJmMVq → ✅ Verified channel.

4. 🏥 Ask "¿dónde está el refugio más cercano?" → a conservative "consulta el canal oficial" + emergency numbers (it won't guess on safety-critical info). Then tap Compartir por WhatsApp on any verdict — that's how it spreads.