April 3, 2026
Spill the secret sauce
A Recipe for Steganogravy
Hiding secrets in recipe blogs has commenters cackling — and a little spooked
TLDR: A new tool hides secret messages inside ordinary recipe blog intros, and commenters are both delighted and unnerved. Jokes about “alchemical texts” collided with worries that bots could trade hidden messages, as even AI skeptics applauded the cleverness — and wondered what happens when dinner blogs become covert channels.
The internet is slurping up “steganogravy,” a cheeky tool that hides secret messages inside chatty recipe blog intros — and the comments are the real feast. The project, tbrockman/recipe-blog-encoding, promises to tuck your links or notes into utterly normal-sounding intros about garlic chicken and weeknight dinners. Even an outspoken AI skeptic showed up to tip their hat to the craftsmanship, while joking about teaching Anthropic’s AI “Claude” this new kitchen trick.
Fans are howling at the absurdity and brilliance: one commenter called it a “great way to hide your alchemical texts,” while another deadpanned, “now to decode the blog post’s hidden message.” But the spiciest thread? A commenter ran straight to the scary part: if it sounds like “normal slop,” bots could pass hidden notes to each other without humans noticing. Suddenly, grandma’s casserole intro looks like a spy novel.
The tool’s own tone — “vibe-coded” and cheekily “partially plagiarized” — only added to the meme storm. Some praised the site’s design, others imagined recipe wars between stealthy AIs, and everyone agreed: it’s wild that secret data can hide in something this boring. Whether you see it as art, prank, or future problem, the crowd’s reaction is pure entertainment — and a little nervous laughter.
Key Points
- •tbrockman/recipe-blog-encoding is a Python CLI that hides messages in natural-language text using neural linguistic steganography.
- •The tool targets recipe blog introductions as cover text, generating normal-sounding, SEO-style paragraphs.
- •Encoding and decoding require the same shared prompt and model to reliably recover the hidden message.
- •The implementation follows arithmetic coding steganography, mapping secret bits to token choices based on model probabilities.
- •An example demonstrates encoding a URL with a Qwen3-32B model file and successfully recovering it via the decode command.