From c9bac1754cf7315f7fb53c96322203864db8e753 Mon Sep 17 00:00:00 2001
From: osmarks <osmarks@mycorrhiza>
Date: Wed, 10 Jun 2026 18:09:45 +0000
Subject: [PATCH] =?UTF-8?q?Edit=20=E2=80=98autogollark=E2=80=99?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 autogollark.myco | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/autogollark.myco b/autogollark.myco
index 8a7caec..22313c6 100644
--- a/autogollark.myco
+++ b/autogollark.myco
@@ -37,9 +37,10 @@ Autogollark currently comprises the dataset, the search API server and the [[htt
 * https://arxiv.org/abs/2507.07101
 * https://arxiv.org/abs/2507.01335
 * https://arxiv.org/abs/2510.14901
-* https://github.com/d0rc/egg.c and https://eshyperscale.github.io/. Does this actually work (at scale)? Why? Would be really nice for using AMX units.
+* https://github.com/d0rc/egg.c and https://eshyperscale.github.io/. Does this actually work (at scale)? Why? Would be really nice for using AMX units. But no pretrained models.
 * Maybe compute grants are available for training.
 * Substantial bandwidth bottleneck on CPU (230GB/s nominal; 200GB/s benchmarked; 100GB/s per NUMA node, which llama.cpp handles awfully). Specdec/MTP would be useful. Can anything use AMX well though?
+* https://dnhkng.github.io/posts/rys/
 }
 * Search over conversations with non-gollark simulacra? Should find //something// to use spare parallelism on local inference. Best-of-n? https://arxiv.org/abs/2505.10475
 * {Longer context, mux several channels.