Real or Fake?: A second opinion on the internet’s questionable claims

The internet says a lot of things. Real or Fake? gives questionable posts, messages and screenshots a second look, checking 21 signals from scam tactics to satire. Open Jev’s read to see what the bird picked up.

What real task does your product handle with GPT-6 Astra?

GPT-6 Astra in Codex took this from an idea to a deployed public app. It built the text, link and screenshot workflow, connected backend OCR and Jev, designed the 21-signal result view, and implemented rate limits, a daily spending cap and usage telemetry. It also helped test longer inputs, fix OCR failures and refine the mobile experience. The user task is getting a quick second opinion on dubious posts and messages: spotting scam tactics, exaggeration, missing context or a joke being taken literally, with the judgments visible. Jev supplies those judgments at runtime. GPT-6 Astra was the coding and development agent used to build, test and launch the product.

I’ve been having a lot of fun with Jev and wanted to try something people could play with. The internet provided plenty of material. Real or Fake? looks at questionable posts, messages, links and screenshot text. It checks 21 things, including scam tactics, exaggeration, humor, missing context, and whether a quote sounds like its supposed author. You get a quick take first; open Jev’s read if you want to see the judgments behind it. It can also say “can’t call this one.” I built it with GPT-6 Astra in Codex and hosted it on ChatGPT Sites. Astra helped with the app, design iterations, testing and deployment. Jev runs the checks, with a small Cloudflare worker reading the text in screenshots. It’s free to try, with limits to keep the bird’s running costs under control. I fed it Jev’s own launch post. It wasn’t entirely sold. If you find something it gets wrong, I’d love to see it.

What real task does your product handle with GPT-6 Astra?