README: community-first intro

README.md

@@ -14,12 +14,17 @@ tags:
 
 # GLM-5.3-Flash + DFlash2 on 2× NVIDIA DGX Spark (GB10) — SGLang TP=2 recipe
 
-First published deployment of [incoai/GLM-5.3-Flash-DFlash2](https://huggingface.co/incoai/GLM-5.3-Flash-DFlash2)
-drafting for [LibertAIDAI/GLM-5.3-Flash-NVFP4](https://huggingface.co/LibertAIDAI/GLM-5.3-Flash-NVFP4)
-on the upstream **SGLang** path (PR [#36507](https://github.com/sgl-project/sglang/pull/36507) branch,
-incl. #36708), on consumer Blackwell (sm_121).
+Hey — I'm one of the folks running DGX Sparks at home, and this community's recipes are
+the only reason my cluster works at all. tonyd2wild's GB10 forensics, MiaAI-Lab's
+dual-Spark configs, hasso5703's DFlash2 writeup, LibertAIDAI's quant card — I've leaned
+on all of them, so here's mine back.
 
-**Source of truth & full recipe:** https://github.com/beastllama/GLM-5.3-Flash-DFlash2-SGLang-2x-DGX-Spark
+This is GLM-5.3-Flash with the incoai DFlash2 drafter on the **SGLang** path (the
+PR [#36507](https://github.com/sgl-project/sglang/pull/36507) branch everyone will get
+by default once it merges). Getting it to boot on GB10 took a night and four fixes
+nobody had written down yet — they're all here with patches and probes, so your
+bring-up should take an hour instead. If you hit something new, open an issue and
+I'll dig in with you.
 
 ## Measured (warmed, temp 0, stream:false, 800 tok, n=5 medians, stock clocks)
 | prompt | DFlash2 ON | no-spec same stack | speedup |