build-small-hackathon/heuremen-fallback-corpus
Heuremen Fallback Corpus 301 handwritten prompt-response pairs across 22 apps, each crafted by hand to sound like a friend, not a professor. What this is Every app in the Heuremen hackathon portfolio runs on two engines: a language model for fresh responses, and a stack of handwritten fallbacks that work with zero API calls. This dataset is the fallback stack — the responses that ARE the product, not a safety net. Why it's valuable Most datasets are… See the full description on the dataset page: https://huggingface.co/datasets/build-small-hackathon/heuremen-fallback-corpus.
Heuremen Fallback Corpus
301 handwritten prompt-response pairs across 22 apps, each crafted by hand to sound like a friend, not a professor.
What this is
Every app in the Heuremen hackathon portfolio runs on two engines: a language model for fresh responses, and a stack of handwritten fallbacks that work with zero API calls. This dataset is the fallback stack — the responses that ARE the product, not a safety net.
Why it's valuable
Most datasets are scraped, generated, or crowdsourced. This one was written by hand at 5am on spite and breakfast burritos. Every entry was written to meet specific constraints:
- Sound like a friend, not a professor
- No jargon, no "did you know," no lecture energy
- Be specific but not technical
- Be warm but not sappy
- Make people see something they already knew but hadn't noticed
Categories
Format
JSON array. Each entry:
{
"app": "myth-unpacker",
"category": "fallbacks",
"topic": "knocking on wood",
"response": "This one goes back to a time when...",
"system_prompt": "Someone names a myth..."
}Usage
Use as few-shot examples, fine-tuning data for empathetic response generation, or as a reference for prompt engineering that prioritizes warmth over accuracy-signaling.
License
MIT. Use it however you want. Just don't make it sound like a professor.
Heuremen — Let's stay connected and keep information free.