Update README.md

README.md

@@ -13,9 +13,12 @@ tags:
   - build-small-hackathon
   - thousand-token-wood
   - tiny-titan
+  - off-the-grid
   - off-brand
   - best-agent
   - best-demo
+  - field-notes
+  - sharing-is-caring
   - simulation
   - creative-writing
   - gradio
@@ -29,6 +32,11 @@ tags:
 
 Built for the **Build Small Hackathon 2026 — Thousand Token Wood** track.
 
+**Demo video:** https://youtu.be/UPDATE_THIS_LINK  
+**Social post:** https://x.com/iagained/status/2066384319929778645?s=20  
+**Blog post:** https://huggingface.co/blog/nishtha711/tiny-civilization  
+**Agent traces:** https://huggingface.co/datasets/build-small-hackathon/tiny-civilization-traces
+
 ---
 
 ## What is this?
@@ -51,27 +59,40 @@ Then a fifth **Narrator agent** writes a formal newspaper front page treating th
 | Feature | Detail |
 |---|---|
 | 🗞️ **Animated Newspaper** | Paper drops in, headline types itself, ink fades in |
-| 🔊 **Read Aloud** | Browser TTS reads today's front page |
-| 🎵 **Sound Effects** | Paper/press sounds via Web Audio API |
+| 🔊 **Read Aloud** | Browser TTS reads today's front page — no API needed |
+| 🎵 **Sound Effects** | Paper/press sounds via Web Audio API (synthesised) |
 | 🐿️ **Running Squirrel** | Occasionally sprints across the bottom of the screen |
 | ⏳ **Loading Messages** | Rotating funny messages while the model runs |
 | 📜 **Archive** | Browse every past front page |
-| 🗣️ **Spread a Rumour** | Influence the next day's events |
+| 🗣️ **Spread a Rumour** | Pick a creature + rumour type — affects the next day |
 | 🎁 **Donate a Weird Object** | Items enter the creature economy |
 | ⚖️ **Propose a Law** | Shift the legal framework of the Hollow |
 | 🖼️ **Share as Image** | Download a PNG of today's front page |
 | 🎮 **Konami Code** | `↑↑↓↓←→←→BA` — reveals all agent system prompts |
-| 💾 **Persistent SQLite** | Civilisation remembers its history across sessions |
+| 💾 **Persistent SQLite** | Civilisation remembers history across sessions |
 
 ---
 
 ## Model
 
-- **Primary**: `Qwen/Qwen2.5-1.5B-Instruct` — 1.5B parameters
-- **Fallback**: `Qwen/Qwen2.5-3B-Instruct` — 3B parameters
-- **Both qualify** for the ≤4B **Tiny Titan** badge 🐜
-- Runs locally via **ZeroGPU** — no external AI APIs 🔌
-- ~520 tokens per day across 4 inference calls
+- **Primary**: `Qwen/Qwen2.5-1.5B-Instruct` — **1.5B parameters** ≤ 4B → 🐜 Tiny Titan
+- **Fallback**: `Qwen/Qwen2.5-3B-Instruct` — 3B parameters, also ≤ 4B
+- Runs **100% locally** via ZeroGPU — zero external AI APIs 🔌
+- ~520 tokens per day across 4 inference calls (~15 seconds on T4)
+
+---
+
+## Badge Claims
+
+| Badge | Qualification |
+|---|---|
+| 🐜 **Tiny Titan** | Qwen2.5-1.5B = 1.5B params, well under the 4B cap |
+| 🔌 **Off the Grid** | No cloud AI APIs — model runs locally on ZeroGPU |
+| 🎨 **Off-Brand** | Fully custom newspaper UI: UnifrakturMaguntia font, two-column CSS layout, CSS keyframe animations, Web Audio API sounds — zero default Gradio styling visible |
+| 🤖 **Best Agent** | 5 independent agents: Fox, Badger, Squirrel, Mole (distinct system prompts + relationship scores), plus Narrator |
+| 🎬 **Best Demo** | Demo video + social post + working Space |
+| 📡 **Sharing is Caring** | Agent traces exported at [HF Dataset](https://huggingface.co/datasets/build-small-hackathon/tiny-civilization-traces) |
+| 📓 **Field Notes** | Blog post at [HF Blog](https://huggingface.co/blog/nishtha711/tiny-civilization) |
 
 ---
 
@@ -80,23 +101,24 @@ Then a fifth **Narrator agent** writes a formal newspaper front page treating th
 ```
 advance_day()  ←── @spaces.GPU(duration=60)
   │
-  ├─ call_agent("fox", trade_prompt)         → ~80 tokens
-  ├─ call_agent("badger", gossip_prompt)     → ~80 tokens
-  ├─ call_agent("squirrel", feud_prompt)     → ~80 tokens
-  └─ generate_newspaper(events_summary)      → ~280 tokens
+  ├─ call_agent("fox",      situation_prompt)   → ~80 tokens
+  ├─ call_agent("badger",   situation_prompt)   → ~80 tokens
+  ├─ call_agent("squirrel", situation_prompt)   → ~80 tokens
+  └─ generate_newspaper(events_summary)         → ~280 tokens
          │
-         └─ Structured output: HEADLINE / article / WEATHER / FOX / BADGER / SQUIRREL / MOLE
+         └─ Structured output:
+            HEADLINE / article / WEATHER / FOX / BADGER / SQUIRREL / MOLE
 ```
 
-Each agent has a 4-sentence system prompt defining their personality. The narrator uses a rigid output schema that the 1.5B model follows ~85% of the time, with graceful fallbacks for the rest.
+Each agent has a 4-sentence system prompt defining their personality. The narrator uses a rigid output schema with a worked example (few-shot) so the 1.5B model follows the format ~90% of the time, with graceful fallbacks for the rest.
 
-Creature relationships drift based on event type (+5 trade, +7 invention, -4 gossip, -10 feud) and feed back into future agent prompts, creating emergent character dynamics over time.
+Creature relationships drift based on event type (+5 trade, +7 invention, −4 gossip, −10 feud) and feed back into future agent prompts — emergent character dynamics over time.
 
 ---
 
 ## Persistent History
 
-Everything is stored in **SQLite**: days, events, creature relationships, inventory, nudges. The newspaper prompt includes the last two published headlines, so the model can write "In a development reminiscent of last Tuesday's acorn scandal..." History actually matters.
+Everything is stored in **SQLite**: days, events, creature relationships, inventory, nudges. The newspaper prompt includes the last two published headlines so the model writes *"In a development reminiscent of last Tuesday's acorn scandal..."* — history actually matters.
 
 ---
 
@@ -113,9 +135,9 @@ CREATE TABLE nudges   (id, day_number, nudge_type, nudge_value);
 
 ## Easter Egg
 
-Type the **Konami Code** while the app is focused:  `↑ ↑ ↓ ↓ ← → ← → B A`
+Type the **Konami Code** while the app is focused: `↑ ↑ ↓ ↓ ← → ← → B A`
 
-A modal appears showing all five raw system prompts — the behind-the-scenes look at what's driving the simulation.
+A modal appears with all five raw system prompts — the behind-the-scenes look at what drives the simulation. Only the woodland elite know this.
 
 ---