backlog: Add Wish List section with Task #96 (Gemma 4)

New section: 🌟 WISH LIST — Make It Happen Someday

Task #96: Gemma 4 Self-Hosted LLM
- Deploy Gemma 4 26B A4B (MoE) on TX1 via Ollama
- Zero API costs for Trinity Codex
- TX1 has 251GB RAM, perfect for CPU-only MoE inference
- Links to Gemini consultation document

Dreams we'd love to make real when the stars align.

Signed-off-by: Claude (Chronicler #63 - The Pathmaker) <claude@firefrostgaming.com>

BACKLOG.md

@@ -1,6 +1,6 @@
 # 📋 BACKLOG — Prioritized Task List
 
-**Last Updated:** April 6, 2026 by Chronicler #62  
+**Last Updated:** April 6, 2026 by Chronicler #63  
 **Purpose:** Organized, prioritized collection of active tasks
 
 ---
@@ -78,6 +78,24 @@ Personal projects and Trinity member tasks.
 
 ---
 
+## 🌟 WISH LIST — Make It Happen Someday
+
+Dreams we'd love to make real when the stars align.
+
+| # | Task | Why It's Amazing | What We Need |
+|---|------|------------------|--------------|
+| 96 | Gemma 4 Self-Hosted LLM | Zero API costs, frontier AI for Trinity Codex, data never leaves TX1 | Time to test on TX1, validate t/s speed |
+
+**Task #96 Details:**
+- **What:** Deploy Gemma 4 26B A4B (MoE) on TX1 via Ollama
+- **Why:** Replaces $50-200/month in API costs with $0 self-hosted inference
+- **Hardware:** TX1 has 251GB RAM, CPU-only — perfect for MoE architecture
+- **Integration:** Connect to Dify for Trinity Codex RAG
+- **Commands:** `ollama pull gemma4:26b-a4b-q8_0` (~26GB)
+- **Consultation:** `docs/consultations/gemini-gemma4-selfhosting-2026-04-06.md`
+
+---
+
 ## ✅ RECENTLY COMPLETED
 
 | # | Task | Completed | By |
@@ -118,7 +136,8 @@ These tasks are no longer relevant due to architecture changes:
 | LOW | 11 |
 | PERSONAL | 7 |
 | BLOCKED | 1 |
-| **TOTAL ACTIVE** | **30** |
+| WISH LIST | 1 |
+| **TOTAL ACTIVE** | **31** |
 
 ---