[Built by Snigdha Babbar — LinkedIn, ⏱ 6 min read]

Live demo → https://whatsapp-forward-checker-rust.vercel.app

See it vs ChatGPT → https://whatsapp-forward-checker-rust.vercel.app/compare

The moment that started it

"I keep getting forwards from my parents and uncles that I'm pretty sure are fake, but I have no fast, respectful way to push back without starting a family argument or sounding like a know-it-all."

That's the user. It's also me.

Every working professional in India is in 4–6 family and community WhatsApp groups, getting 5–10 forwards a week from older relatives — at least half of them suspicious. The current options are all bad:

• Ignore it → the misinformation spreads.
• Reply "this is fake" → family fight.
• Spend 20 minutes Googling → by then the forward has been re-shared ten times.
• Ask ChatGPT → get a paragraph that doesn't quote the message, doesn't suggest a polite reply, and isn't shaped for sharing back to the group.

The real pain isn't "is this true?" — it's "how do I respond to my dad without it becoming a thing?" That distinction shaped everything I built.

The reframe that unlocked the product

I had one day. Real fact-checking needs a curated source corpus, embeddings, a vector DB, and a freshness pipeline — none of which I could build in 24 hours.

So I stopped trying to verify truth.

Instead, I built around what an LLM is genuinely great at: detecting manipulation patterns. Fake forwards have a tell — fabricated urgency, false authority, recycled news, unsourced numbers, translation artifacts. An LLM can spot these from the text alone. No retrieval needed.