@chase_ai_: Karpathy's Obsidian "RAG"

Chase AI
Chase AI
Open In TikTok:
Region: US
Saturday 04 April 2026 21:57:27 GMT
56310
2701
41
720

Music

Download

Comments

_orange_triangle
🟪 Orange Triangle :
Quit pushing obsidian
2026-04-05 21:17:51
3
e___m___k
Evgeny :
you don't need Obsidian for this IMHO. Claude Code can just manage the file structure directly. It's basically. Good idea, but kinda obvious one
2026-04-05 05:53:29
5
lucymangin1
lucymangin1 :
We ended up using Supabase to design a database as the OS in our business, I’d like to have an obsidian layer that communicates the strategies & workflows that references the database. Could I have Supabase as my vault? I don’t want to have duplicate info
2026-04-14 19:43:27
1
ahoppity
Aaaaah :
lol, large amounts? Nah. There’s a huge difference between a bunch of personal knowledge .md files on your desktop and a 50tb vector database of knowledge documents.
2026-04-15 14:42:57
2
mbensingercolton
Mikey Colton :
do you know how it chunks the content and how detailed it is in content retrieval?
2026-04-08 00:11:31
3
seerproductionss
Dave C :
Haha but what about your Obsidian vault *does* have embeddings and syncs to Memgraph, 😂
2026-04-24 07:06:42
1
jzsn31
jzsn31 :
You’re one of the best content creators on the platform. Thanks for doing this!
2026-04-04 23:50:20
10
_v.i.c._
Victor :
Augmentation is used loosely
2026-04-05 03:32:36
2
chestercoonen
chestercoonen :
still too slow. we need to focus on how the etcd works
2026-04-07 16:47:02
2
wyrobotic
wyrobotic :
I switched from notion and congee to obsidian. And there are more reasons to do this than just simplicity most people don’t realize that if you use something like notion that it consumes a tremendous amount of your context window by referencing your entire notion database instead of a simple flat folder of markdown files. It is the easiest way to extend acknowledge base
2026-04-05 05:38:53
4
stein_247
Jack :
Would this work for data engineering / notebook execution / coding
2026-04-10 05:15:10
2
jacobshipss
Jake | AI :
Great for solo operators!
2026-04-06 08:47:56
2
channel.hoa.k.usa
channel hoa kỳ USA :
how you teach me AI make coins
2026-05-24 15:28:19
2
x3h013
X3H :
Karpathy designed LLM Wiki for one person. I scaled it to 4 servers and 7 apps — Claude Code sessions sync with each other through git, without ever talking directly. github.com/alinclaudiu/vaultmesh
2026-05-09 18:48:34
1
nathanielplzz
nathanielplzz :
why did you turn off comments on the video before this?
2026-04-06 02:53:17
2
ozais5
Ozais :
im to newbie for this 😅 I dont know what a Rag is. still learning how to vibe code lol
2026-04-07 06:50:26
2
zach_aryy
Zach :
You rock dude
2026-04-05 14:11:52
2
ramdomize0
BigR :
Why though
2026-04-05 14:43:44
2
matasovic
Matasovic :
compared to hermes?
2026-05-28 16:56:51
0
ronny_lutz
ronny_lutz :
Do you Think its a good Option for a small Team (20 people) i configure it and we use it for all Internal Informations. Is it possible to Host it Self oder in Europe?
2026-04-06 17:11:23
0
bedfordharbour
Ryan Joyce :
I set up a wiki a week after Karpathy's post and have been experimenting with it. It is still prone to hallucination, making connections that don't exist and incorrectly picking and choosing what information to ingest. It's a great concept but there is a large amount of work to do before it can handle a large amount of data.
2026-05-23 04:02:32
0
gymnasiearbete
gymnasie arbete :
💯 Great video
2026-04-04 22:11:25
0
der_ziener
Maxi Ziener :
rag
2026-05-05 21:51:55
0
kaxf_
Kaxf :
people were doing this last year and now it’s finally mainstream lol
2026-04-05 00:58:31
6
loloquan
LoloQuan :
Wouldn’t notion be the same thing?
2026-04-04 23:56:44
4
To see more videos from user @chase_ai_, please go to the Tikwm homepage.

Other Videos


About