README/write-up: surface demo CTA, claim Tiny Titan, link field notes

Address the friendly-review feedback: promote the live-demo link to a
prominent CTA under the title, foreground the Tiny Titan (<=4B) claim in
the badge line and merit-badges section, and add a closing field-notes
pointer. Launch-post link confirmed correct (review's "broken" flag was
an X bot-wall false positive).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

README.md

@@ -20,9 +20,11 @@ tags:
 
 **Dump everything on your mind — get back what to do *first*, with the math shown.**
 
-[**▶ Live demo**](https://huggingface.co/spaces/build-small-hackathon/whatfirst-small)  ·  [📣 Launch post](https://x.com/tbd_ntbd/status/2066222870657692128)  ·  [what-first.com](https://what-first.com)  ·  Apache-2.0
+### [▶ Try the live demo](https://huggingface.co/spaces/build-small-hackathon/whatfirst-small)
 
-🤏 **3B params** (≤ 4B)  ·  🔌 **runs 100% offline** — no internet required  ·  🦙 llama.cpp
+🤏 **3B params** (≤ 4B — *Tiny Titan*)  ·  🔌 **runs 100% offline** — no internet required  ·  🦙 llama.cpp
+
+[📣 Launch post](https://x.com/tbd_ntbd/status/2066222870657692128)  ·  [what-first.com](https://what-first.com)  ·  Apache-2.0
 
 A small **local** vision-language model (Qwen2.5-VL-3B, ~2 GB, running on
 llama.cpp) reads a messy brain-dump or a photo of a to-do list and turns each
@@ -104,3 +106,8 @@ python -m pytest test_score.py    # or: python test_score.py
   (Q4_K_M + f16 mmproj), ≤ 32B and laptop-runnable.
 - **Off the grid:** all inference is local llama.cpp over localhost; nothing
   leaves the box at runtime.
+
+---
+
+📓 **See the full story in our [field notes](submission/whatfirst-small-writeup.md)** —
+what worked, where a 3B model wobbles, and how the design absorbs it.

submission/whatfirst-small-writeup.md

@@ -79,6 +79,8 @@ Disagree with a score? Drag a slider and the list **re-ranks live** — the mode
 
 ## Merit badges
 
+**🤏 Tiny Titan.** The entire app — brain-dump parsing, image reading, and live re-scoring — runs on a single 3B model (Qwen2.5-VL-3B), comfortably under the 4B line, on CPU.
+
 **🔌 Off the Grid.** All inference is local llama.cpp over localhost. Nothing leaves the box at runtime — no cloud, no API keys. Your to-do list, including a photo of it, never travels.
 
 **🦙 Llama Champion.** The whole language stack is a llama.cpp `llama-server`, built from source in the container, serving a Qwen2.5-VL GGUF (text **and** vision) behind an OpenAI-compatible API.