@gojutechtalk: The idea of distilling LLM capabilities into specialized SLMs for local use is gaining traction. This approach trades generalizability for efficiency, allowing users to switch between models based on specific tasks like coding, creative writing, or translation. It's like having a library of video games – you access the one you need, when you need it. Link to full live stream: https://youtube.com/live/xyBHj2b-oNY #AI #LLM #SLM #MachineLearning #TechnologyTrends #FutureofAI
Faster AI router maybe with coupling (separated generalize) weights/Layers between the special AI's. You can connect 1000 AI's with 20-30% connection intelligence
2026-07-06 09:26:18
0
Seth The Teacher Talks :
Oh wow, this is novel, and it could be the next big thing, LLM stripping. Because look, you can literally pause Claude Code mid coding and ask it about Harry Potter, and it will tell you everything about Harry Potter. Why can't I have a Coding specialist who doesnt know harry Potter?
2026-07-05 13:06:44
2
sonicpoet :
I’ve been saying this since day 1
2026-07-03 06:41:59
1
DirtyRottenXer :
Totally agree. I use a llm to orchestrate tasks that are largely made up of automations. It’s unnecessary… but cool to have. If I had a smaller rig, a slm could absolutely accomplish 90% of the work. The best workflow comes from an operator who is good at organization and delegation.
2026-07-03 14:40:27
0
juanitomint :
I'm testing this in local environment with models that fits on 32gb VRAM but would like to have more specialized models
now tinkering with ornith for programming
2026-07-05 23:28:21
0
techwingmanage :
💯💯💯Correct
2026-07-03 01:17:12
0
mrqs :
Small LLMs will work but it will take a lot of years and big LLMs will still be the priority as they’re much smarter and capable
2026-07-04 18:11:44
0
Argentus :
Given an llm a word500 puzzle to see how smart they are!
2026-07-05 15:24:45
0
Cr :
For coding are not that great but for agents routing and tools / MCP usage are perfect, I downgraded my agents workflows from APIs calls to local and realized even qwen 4b was giving great results under 2sec with 10k context, however for deep planing need to switch to 8b or 35b A3B
2026-07-04 02:44:29
0
sharon :
could you kindly tackle RAG?
2026-07-07 21:43:49
0
The Polymath :
Agreed
2026-07-04 09:05:58
0
DiegoV :
distill human + LLM -> specialized slm
2026-07-05 22:53:36
0
Moustafa Abd Elmegied :
🥰🥰🥰
2026-07-05 19:17:09
0
Charlotte Wright :
AI dropshipping encouraged me to become a better learner.
2026-07-20 15:26:28
0
To see more videos from user @gojutechtalk, please go to the Tikwm
homepage.