+- Hack Forums (https://hackforums.net)
+-- Forum: Tech (https://hackforums.net/forumdisplay.php?fid=444)
+--- Forum: Artificial Intelligence (https://hackforums.net/forumdisplay.php?fid=460)
+---- Forum: AI Programming and Vibe Coding (https://hackforums.net/forumdisplay.php?fid=462)
+---- Thread: Compute power= Personal LLM (/showthread.php?tid=6330141)
Compute power= Personal LLM - iconia - Aug 11, 2026
I've been able to configure software that maximizes the compute on your setup without needing higher order silicone, as unbelievable as that may sound.
My main goal has been to focus on the concept of portable competence at the kernel layer. To give everyone the power of a cluster locally, so they can decentralize the llm dependency and capacity.
To my surprise it didn't take as long as I thought it would, and smarter models at that.
That my entire focus for now, bolstering compute for higher complexity and competent machinery.
Hopefully the world is ready for that without a riot, I have a long journey ahead of me with reaching the masses.
RE: Compute power= Personal LLM - ◔ßorder╙ineExtrⒶ - Aug 18, 2026
This is a concept straight out of sci-fi. How were you able to do that, programmatically? I've been looking into overclocking CPUs though my AMD is outdated and I'm mostly worried about an increase in performance which could lead to early degeneration or in other terms burning down the nanowafers on a mollecular level.
I think that recycling computational power should lead to longer term operational efficiency & hardware capabilities that are adaptable to current trends, although lower peak performance when comparing chipset gens. As an added bonus, costs remain the same or even less.
RE: Compute power= Personal LLM - iconia - Aug 18, 2026
(Aug 18, 2026 04:51 AM)◔ßorder╙ineExtrⒶ Wrote: This is a concept straight out of sci-fi. How were you able to do that, programmatically? I've been looking into overclocking CPUs though my AMD is outdated and I'm mostly worried about an increase in performance which could lead to early degeneration or in other terms burning down the nanowafers on a mollecular level.
I think that recycling computational power should lead to longer term operational efficiency & hardware capabilities that are adaptable to current trends, although lower peak performance when comparing chipset gens. As an added bonus, costs remain the same or even less.
In my wheel house for sure and you're on the right path, a lot of my approach came to speeding up matrix math by using pure math and metaphysics approaches to information theory. Once you can change how information is both internalized, metabolized and the like you can change the workloads dramatically. TDA (Topological Data Analysis) which is what Karpathy is focused on is a huge factor conceptually speaking in achieving this. But I cant give away too much yet lol.
RE: Compute power= Personal LLM - Shapeless - Aug 18, 2026
I dont know why but i believe this guy. And if theres money in it i will join your fellowship.
RE: Compute power= Personal LLM - Undead - Aug 18, 2026
![[Image: tenor.gif]](https://c.tenor.com/g_YopdgUyLcAAAAd/tenor.gif)
RE: Compute power= Personal LLM - iconia - Aug 18, 2026
(Aug 18, 2026 01:18 PM)Shapeless Wrote: I dont know why but i believe this guy. And if theres money in it i will join your fellowship.
thanks for the vote of confidence! I wouldn't deceive anyone at this level the people here are too smart to try to pull anything. I will release it when its fully done, there isn't any money involved yet cause its just me, but I don't think I will need any investors once people see what this can do.
(Aug 18, 2026 01:34 PM)Undead Wrote:
i look forward to you using it. it will blow your mind joe