Add hackathon submission tags + idea/how-built/tech + submission link slots

README.md

@@ -9,18 +9,63 @@ app_file: app.py
 pinned: false
 license: apache-2.0
 short_description: UX writing arena — base model vs fine-tune. You judge.
+tags:
+  - backyard-ai
+  - best-demo
+  - tiny-titan
+  - judges-wildcard
+  - modal
 ---
 
 # ⛺ Copy Campfire — the UX writing arena
 
 Two UX writers by the fire — one went to training camp. Paste your UI copy, get two
 blind rewrites (base **Qwen3.6-27B** vs the **ux-writing-1** QLoRA fine-tune), vote,
-then see the reveal. Votes are stored privately as preference data that trains the next version.
+then see the reveal. Votes are stored privately as preference data that trains the next
+version — the arena is the data flywheel.
 
-Built for the Hugging Face **Build Small** hackathon — small models, big adventure.
+## The idea
 
-- Model: [`gr33r/ux-writing-1`](https://huggingface.co/gr33r/ux-writing-1)
-- Code: [`content-designer/ux-writing-1`](https://github.com/content-designer/ux-writing-1)
+Every product ships copy debt — "Invalid", "OK", "An error occurred while processing
+your request" — thousands of interface strings nobody owns. **ux-writing-1** is a small,
+open model fine-tuned to review that copy the way a senior UX writer would: rewrite
+what's weak, **keep what's already right**, and say why. It runs at codebase scale for
+cents, on hardware you control — exactly the kind of bulk, structured LLM work that
+shouldn't be metered at frontier prices.
+
+## How it was built
+
+- **Base:** [Qwen3.6-27B](https://huggingface.co/Qwen/Qwen3.6-27B) — Apache-2.0, 27.8B
+  params (under the 32B limit).
+- **Training:** QLoRA on ~1,400 hand-built before/after pairs, each with a
+  `{rewrite, reason, risk}` contract and the code context the string lives in. Two runs,
+  ≈$30 of GPU on **Modal**.
+- **Serving:** Modal OpenAI-compatible endpoint; this Gradio arena calls it live.
+- **Honest eval:** 83% blinded human preference vs the base model (90 held-out items) —
+  not automated heuristics, which saturated.
+- **At codebase scale:** reviewed 10,000 of PostHog's UI strings in 77 min for **$3.22**
+  — changed 994, **kept 9,006** as-is. Restraint is trained, not hoped for.
+
+## Tech
+
+Qwen3.6-27B · QLoRA (PEFT / TRL) · Modal (training + serving) · Gradio (this arena) ·
+llama.cpp / GGUF (zero-cost laptop inference) · Hugging Face Hub (model, adapter,
+dataset, blinded vote store).
+
+## Take it home
+
+- **Model:** [`gr33r/ux-writing-1`](https://huggingface.co/gr33r/ux-writing-1) ·
+  [adapter](https://huggingface.co/gr33r/ux-writing-1-lora) ·
+  [GGUF](https://huggingface.co/gr33r/ux-writing-1-GGUF)
+- **Code & eval tooling:** [`content-designer/ux-writing-1`](https://github.com/content-designer/ux-writing-1)
+- **Interactive figures:** [the campfire gallery](https://copy-campfire-gallery.vercel.app)
+
+## Submission
+
+- 🎬 **Demo video:** _add the public video link here before submitting_
+- 📣 **Social post:** _add the social post link here before submitting_
+
+---
 
 **Space secrets required:** `BATTLE_URL`, `AUTH_TOKEN` (Modal backend), `HF_TOKEN`
 (write access for vote storage); optional `VOTES_DATASET`.