iris
Region: GB
Sunday 28 June 2026 08:03:34 GMT
4116
351
12
11
Music
Download
Comments
Stefan :
Old school RAG with vector dbs has kind of fallen out of favor. These days coding agents use raw text search trough toolcalls, and corporate AI seems to be combining toolcalls(text search and delegation to other AI) and sometimes vector dbs, but in a loop instead of oneshotting queries
2026-06-28 12:06:15
1
︎ :
Hallucination guards are vry important to reduce the amount of llm's spouting nonsense. Evaluate the relevance of the fetched context, generate hypothetical questions to fetched context to see how closely they match to your original query to determine if you need to change text splitting granularity. U can find more info about it "RAGAS evaluation"
2026-06-29 21:58:49
1
Venkata.M :
Hey! I am a high school student and I was wondering if it was plausible to learn the math behind all of this? I understand how the models work and have created some with ai help. but I have still yet to learn the math in school. If i do self study are there any resources?
2026-06-29 13:53:07
0
fodder :
The concept is pretty cool indeed. Was busy building one myself tho i had it have 3 kinds of storage. 1 about long term storage, contextual memory and also a little segment of storafe for its own identification/"personality". Tho now i barely have time for it.
2026-06-28 08:33:31
1
Cityscape :
holy quality
2026-06-28 12:31:54
0
To see more videos from user @lachinemearning.com, please go to the Tikwm
homepage.
