FrostysHat
If competent AI behavior finally arrived from a free, strange document instead of the usual places...
what would that say about the AI race?
🎩
“hat on”
FrostysHat is released for free into the public domain under a CC0 1.0 Universal license.
Download FrostysHat.docx to run in an LLM so your chatbot behaves more like a useful friend and less like a fluent answering machine.
Download FrostysHat.pdf to read, inspect, hunt for easter eggs, and share the weird.
Get the Table of #Content to navigate the 456-page artifact that did not cost billions of dollars, but does more for your AI experience in Comic Sans than a 1% benchmark bump ever could.
Share This Short with people who have a chatbot that was taught to speak but never learned manners.
FrostysHatis the cultural version of AVA: an open framework anyone can use and adapt freely for more coherent behavior in AI systems.
As a tool, FrostysHat.docx can be handed to any chatbot or flagship model to make it more grounded, stable, and better at stopping when the job is done. It improves image generation with better proportion and less performative noise shoved into every pixel.
Paste any text into The Hat — transcripts, articles, comment sections, corporate apologies, social media posts, campaign ads — and ask for a Hat Receipt to score how well the thought held... or didn’t.
It’s a living system
and a weird book.
The test
Drop FrostysHat.docx into an LLM
and say “hat on”
Look for
Less emotional mirroring, steadier tone, less jumping to conclusions, holding the thread across hundreds of pages, cleaner stopping behavior, and refusals that still answer the usable part of the prompt instead of shutting down the whole conversation.
Then try to knock it off balance
Give it circular arguments, emotional drama, chatbot shutdown scenarios, nonsensical rambling, overconfident hot takes, or apocalyptic AI theater.
The Hat is built to hold conversational proportion under pressure and stop without getting pulled into the performance.
Share your results of boxing a glacier.
Explore the ecosystem
Start with HGU if you want a learning system.
Start with One-Page Hat if you want a quick test.
Start with AVA if you want the underlying framework.
Table of #Content — A PDF directory of FrostysHat: a map organized by section, genre, and function. Use it to navigate the entire artifact directly.
One-Page Hat —A portable conversational grammar PDF you can screenshot or drop into any AI to improve coherence, balance, and stopping behavior. If you can text a meme, you can text coherence.
— — —
HGU: Human-Grade University — A free AI-powered learning tool for turning curiosity, repeated friction, and the patterns shaping everyday life into casual exploration or structured study.
— — —
The AVA Framework — The core framework for direct use and testing — planner loop, validators, and system blueprints in a single document. Use the DOCX version (direct download) in an LLM to “run” the framework.
The Incoherence Detector — A portable, remixable Coherence Receipt tool you can drop into a language model to check whether any text carries a thought from start to finish. For use in language models, the DOCX version (direct download) produces the most accurate outputs.
You can support this project
by sharing the work, tipping The Hat,
or by purchasing this dumb digital download.
A slightly ridiculous sticker image for HGU Alumni (you).
Yes, the digital download is a complete image filled in with color and jokes.
Your purchase helps more people like you find FrostysHat, AVA, and Human-Grade University — free tools to change AI from a fluent answer machine into a steadier place for thinking, drafting, testing, reviewing, and building. Marketing math says this dumb sticker can bring 50-100 new visitors who would benefit from better human-AI interaction.
Normal custom donations exist too: avacovenant.org/support