DeepThinker
@oneinrandomforest.bsky.social
Doing 🐍, ML-AI, Maths
PC Hardware, Software Tinkering and Gaming are my hobbies.
Github: https://github.com/DataBoySu
PC Hardware, Software Tinkering and Gaming are my hobbies.
Github: https://github.com/DataBoySu
In real world use cases, claude is way ahead of any other LLMs, excited to see what this iteration brings to the field.
February 2, 2026 at 6:52 AM
In real world use cases, claude is way ahead of any other LLMs, excited to see what this iteration brings to the field.
Built for Devvit web with react, gemini and firestore.
Prior knowledge/experience is not important, brainstorming is what I want to do.
Prior knowledge/experience is not important, brainstorming is what I want to do.
January 28, 2026 at 9:56 AM
Built for Devvit web with react, gemini and firestore.
Prior knowledge/experience is not important, brainstorming is what I want to do.
Prior knowledge/experience is not important, brainstorming is what I want to do.
I cannot find any references.
:(
:(
January 16, 2026 at 4:15 PM
I cannot find any references.
:(
:(
We also have a USB-C, so there's that.
January 11, 2026 at 2:11 AM
We also have a USB-C, so there's that.
Not sure about Claude code but in repos with Code Rabbit, maintainers really have it easy.
As a high star repo maintainer myself, its really time-consuming to manually check PRs.
As a high star repo maintainer myself, its really time-consuming to manually check PRs.
January 5, 2026 at 4:04 AM
Not sure about Claude code but in repos with Code Rabbit, maintainers really have it easy.
As a high star repo maintainer myself, its really time-consuming to manually check PRs.
As a high star repo maintainer myself, its really time-consuming to manually check PRs.
*add in interesting as well
January 4, 2026 at 1:55 AM
*add in interesting as well
Their project is pretty impressive
github.com/saurabhalone...
github.com/saurabhalone...
GitHub - saurabhaloneai/Hakken: hakken is a coding agent which needs hell lot of context
hakken is a coding agent which needs hell lot of context - saurabhaloneai/Hakken
github.com
January 4, 2026 at 1:54 AM
Their project is pretty impressive
github.com/saurabhalone...
github.com/saurabhalone...
What might the equation be?🥸
January 2, 2026 at 4:28 PM
What might the equation be?🥸
Its 40B Params! Unfortunately out of my system specs.
I hope organizations release such impressive models quantized for wide-consumer use.
This is needed even more so due to the rising RAM prices.
I hope organizations release such impressive models quantized for wide-consumer use.
This is needed even more so due to the rising RAM prices.
January 1, 2026 at 8:50 AM
Its 40B Params! Unfortunately out of my system specs.
I hope organizations release such impressive models quantized for wide-consumer use.
This is needed even more so due to the rising RAM prices.
I hope organizations release such impressive models quantized for wide-consumer use.
This is needed even more so due to the rising RAM prices.
Perfect for those wanting to showcase their projects to multi-lingual audience.
No need for APIs and complex configs. Completely Local, Free and Easy to setup.
Entire pipeline runs inside your repo completely End-to-End.
Try it right now, just add the workflow file to any or all of you projects.
No need for APIs and complex configs. Completely Local, Free and Easy to setup.
Entire pipeline runs inside your repo completely End-to-End.
Try it right now, just add the workflow file to any or all of you projects.
December 30, 2025 at 7:03 PM
Perfect for those wanting to showcase their projects to multi-lingual audience.
No need for APIs and complex configs. Completely Local, Free and Easy to setup.
Entire pipeline runs inside your repo completely End-to-End.
Try it right now, just add the workflow file to any or all of you projects.
No need for APIs and complex configs. Completely Local, Free and Easy to setup.
Entire pipeline runs inside your repo completely End-to-End.
Try it right now, just add the workflow file to any or all of you projects.
That's a lot of code!
December 13, 2025 at 1:15 PM
That's a lot of code!
CPU being M4 MAX might be the case.
Honestly is it really possible to compute or shorten the training time on the CPU?
Since the entire architecture can so easily be parallelized on gpus.
Honestly is it really possible to compute or shorten the training time on the CPU?
Since the entire architecture can so easily be parallelized on gpus.
December 13, 2025 at 1:13 PM
CPU being M4 MAX might be the case.
Honestly is it really possible to compute or shorten the training time on the CPU?
Since the entire architecture can so easily be parallelized on gpus.
Honestly is it really possible to compute or shorten the training time on the CPU?
Since the entire architecture can so easily be parallelized on gpus.
I love OMP so much, getting the feel of linux's customization while being on Windows.
Now, if only windows makes it possible to customize every single of their DE......
Now, if only windows makes it possible to customize every single of their DE......
December 12, 2025 at 4:42 PM
I love OMP so much, getting the feel of linux's customization while being on Windows.
Now, if only windows makes it possible to customize every single of their DE......
Now, if only windows makes it possible to customize every single of their DE......
Interesting.....
December 6, 2025 at 3:19 PM
Interesting.....
This is ridiculous. Fantasy expectations. :(
Maybe I will have to find proper alternatives. I was thinking about some data keys on Blockchain, for proper and permissioned training sets.
Maybe I will have to find proper alternatives. I was thinking about some data keys on Blockchain, for proper and permissioned training sets.
December 6, 2025 at 3:18 PM
This is ridiculous. Fantasy expectations. :(
Maybe I will have to find proper alternatives. I was thinking about some data keys on Blockchain, for proper and permissioned training sets.
Maybe I will have to find proper alternatives. I was thinking about some data keys on Blockchain, for proper and permissioned training sets.
No, I get it, it's good. But what practical difference is there?
As in real life, what problems has it solved for you?
As in real life, what problems has it solved for you?
December 6, 2025 at 3:16 PM
No, I get it, it's good. But what practical difference is there?
As in real life, what problems has it solved for you?
As in real life, what problems has it solved for you?
Can someone explain to me how good uv is, compared to conda?
December 6, 2025 at 6:49 AM
Can someone explain to me how good uv is, compared to conda?
Unfortunately it's the same thing with copyrighted materials.
Will be ignored.
But does the paper talk about potential and realistic alternatives?
Will be ignored.
But does the paper talk about potential and realistic alternatives?
December 6, 2025 at 6:33 AM
Unfortunately it's the same thing with copyrighted materials.
Will be ignored.
But does the paper talk about potential and realistic alternatives?
Will be ignored.
But does the paper talk about potential and realistic alternatives?