From 74b566dbd24a66319793a7fb7e9c402702d69373 Mon Sep 17 00:00:00 2001
From: osmarks <osmarks@mycorrhiza>
Date: Sun, 20 Apr 2025 20:03:40 +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 | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/autogollark.myco b/autogollark.myco
index 240b617..f3df5fc 100644
--- a/autogollark.myco
+++ b/autogollark.myco
@@ -20,7 +20,9 @@ Autogollark currently comprises the dataset, the search API server and the [[htt
 * Use cheap classifier to evaluate when to respond.
 * Should also allow unprompted messages somehow (polling, rerun after last message?).
 }
-* Tool capabilities (how to get the data? Examples in context only?!).
+* {Tool capabilities (how to get the data? Examples in context only?!).
+* Synthetic via instruct model.
+}
 * {Local finetune only? Would be more tonally consistent but dumber, I think.
 * Temporary bursts of hypercompetence enabled by powerful base model are a key feature. Small model is really repetitive.
 * Can additionally finetune on "interesting" blog posts etc (ref https://x.com/QiaochuYuan/status/1913382597381767471).