Fix submission tags: namespaced track:/sponsor:/achievement: per Build Small validator; add Best MiniCPM Build target

README.md

@@ -9,21 +9,21 @@ app_file: app.py
 pinned: false
 license: mit
 tags:
+  # ── Build Small validator namespaces — ONLY track:/sponsor:/achievement: are
+  #    read by the official README validator (src/lib/readme.ts). Everything
+  #    below them is plain discoverability metadata it ignores. ──────────────
+  - track:wood            # 🍄 Thousand Token Wood — our track
+  - sponsor:modal         # Best Use of Modal — all GPU inference + .map() pre-gen run on Modal
+  - sponsor:openbmb       # Best MiniCPM Build — MiniCPM4.1-8B is the witness's brain (+ VoxCPM2 voice)
+  - achievement:offbrand  # custom courtroom UI, well past stock Gradio
+  # ── discoverability only (not counted by the validator) ──────────────────
   - build-small-hackathon
-  # track (both spellings, per the field guide's note on tag variants)
-  - thousand-token-wood
-  - thousand token wood
-  - adventure-in-thousand-token-wood
-  # sponsor / bonus targets
-  - best-use-of-modal
-  - best use of modal
-  - modal
-  - openbmb
+  - gradio
   - minicpm
   - voxcpm
-  - nemotron
-  - best-agent
-  - best-demo
+  - modal
+  - voice
+  - game
 ---
 
 # ⚖️ WitnessBox — cross-examine a hostile AI witness with your *voice*
@@ -32,7 +32,7 @@ tags:
 > — sound confident and he clams up; sound hesitant and he gets cocky and
 > overshares. Surface **three contradictions** and his voice **cracks** as he breaks.
 >
-> **Track:** 🍄 An Adventure in Thousand Token Wood · **Primary target:** Best Use of Modal
+> **Track:** 🍄 An Adventure in Thousand Token Wood · **Targeting:** Best Use of Modal + Best MiniCPM Build
 
 ---
 
@@ -79,7 +79,8 @@ used as a *platform*, not just a host (the prize counts "inference… all"):
 3. **Parallel `.map()`** pre-generates every scripted beat at deploy time, fanning
    the **32 voice-crack takes across containers at once** and keeping the best.
 4. **Volume** persists the designed CFO reference voice + model cache + chosen beats.
-5. **Memory snapshots** cut CPU-side init on cold start.
+5. **Right-sized GPUs** — an A100 only for the 8B witness brain; the 2B voice and
+   the ASR ride cheaper A10Gs.
 
 **Measured (warm, this deploy).** A live dynamic turn is `MiniCPM4.1-8B` **→ 5.3s**
 for the witness's reply, then `VoxCPM2` **→ 8.6s** for ~4.5s of 48 kHz speech
@@ -90,6 +91,15 @@ by the parallel `.map()` pass and served straight from the Volume, so every
 `$0` via `scaledown_window`. (Container-seconds / $-per-match read live from the
 Modal dashboard, not fabricated.)
 
+## 🧠 Best MiniCPM Build
+The witness *is* a MiniCPM model. `openbmb/MiniCPM4.1-8B` runs the entire persona —
+it reads the delivery stance, decides what Reid admits or hides, and leaks a thread
+toward an uncaught lie when you sound unsure — and `openbmb/VoxCPM2` gives him the
+voice that cracks on the break. The 8B brain is the **core of the experience, not a
+bolt-on**: every line Reid speaks is MiniCPM under a stance- and tier-conditioned
+system prompt, so the drama lives or dies on how well a small model holds a character
+under pressure.
+
 ## Run it
 **Offline (no GPU, no Modal — boots anywhere):**
 ```bash