AI Simulated OS Is Absurd
Related videos
How DeepSeek V4 Broke AI’s Cost Curse
bycloud
101.8k views
What Is Yann LeCun Cooking? JEPA Explained Simply
bycloud
50.7k views
Google's TurboQuant Is Way Too Overhyped
bycloud
20.2k views
The Death of RAG?
bycloud
15.0k views
The RL Irony in LLMs
bycloud
23.0k views
The Chinese AI Iceberg
bycloud
107.4k views
Is It EVEN Possible To Reverse Engineer AI’s Training Data?
bycloud
39.5k views
Anthropic Just Proved Reasoning AIs Would Silently Cheat
bycloud
35.1k views
Gemini 2.5 Pro is just the best choice for AI right now
bycloud
46.9k views
SURREAL Alternate Dimensions & Simulations | The Proof Is Out There
HISTORY
156.8k views
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Unlock all features
FREE: Get instant access to 10 AI summaries, chats, or transcripts per day.
Related videos
How DeepSeek V4 Broke AI’s Cost Curse
bycloud
101.8k views
What Is Yann LeCun Cooking? JEPA Explained Simply
bycloud
50.7k views
Google's TurboQuant Is Way Too Overhyped
bycloud
20.2k views
The Death of RAG?
bycloud
15.0k views
The RL Irony in LLMs
bycloud
23.0k views
The Chinese AI Iceberg
bycloud
107.4k views
Is It EVEN Possible To Reverse Engineer AI’s Training Data?
bycloud
39.5k views
Anthropic Just Proved Reasoning AIs Would Silently Cheat
bycloud
35.1k views
Gemini 2.5 Pro is just the best choice for AI right now
bycloud
46.9k views
SURREAL Alternate Dimensions & Simulations | The Proof Is Out There
HISTORY
156.8k views
Top Comments (10)
It's need to be said that perhaps the most useful "feature" of programs (OS included) is the fact they are deterministic. We have that implicit understanding that "machine is certain" which falls apart when you're replacing them with neural networks - they are anything but certain. And certainty is absolutely critical for OS.
Man I can't wait to use the full power of a 5090 to generate frames of me opening my file explorer!
We have reached a new level of stupid.
Instead of 50mb I shall rent a datacenter for 50 million in order to train on an app I already have!
The Linux world is moving towards more stable, more predictable systems with atomic and declarative distributions, OCI, etc. meanwhile AI companies are like "but what if we went the other direction?"
"I have a PhD in hammerology and this is a nail"
I think rather than the entire OS being built by AI, this kind of research can evolve to make agents understand the OS better.
If it ever becomes a reality, we should call it the Bloat System (or BS for short). Its initialization process should be called the Bloat Loader, its memory, the B-RAM (when the bull meets the ram), its storage is the Slop Disk, and its programming language (prompts), the B#ъ (BShard) language, used to create Bь (BSoft) programs. Uninstalling it is called debloating your PC.
OSes behave the way they do because of the interaction of a large number of files, most of which are not visible on screen. To make a truly neural OS, one would have to train the model to represent the entire OS in its latent space. Not doing that leads to exactly what was shown in the video. There are older, hybrid ideas, such as AIOS, which is better because it doesn't try to replace the highly optimized architecture with an inefficient (in this case) approximation.
me blowing up the imaginary operating system by opening 6 terminal emulators and running dmesg on loop on every one of them
Unlock the Data Inside
Turn Videos into Knowledge
- Get FREE 10/day: transcripts, summaries, chats
- Chat with videos, export text & PDF
- $1 free API credit for RAG, chatbots & research
Free forever plan • All features unlocked
Top Comments (10)
It's need to be said that perhaps the most useful "feature" of programs (OS included) is the fact they are deterministic. We have that implicit understanding that "machine is certain" which falls apart when you're replacing them with neural networks - they are anything but certain. And certainty is absolutely critical for OS.
Man I can't wait to use the full power of a 5090 to generate frames of me opening my file explorer!
We have reached a new level of stupid.
Instead of 50mb I shall rent a datacenter for 50 million in order to train on an app I already have!
The Linux world is moving towards more stable, more predictable systems with atomic and declarative distributions, OCI, etc. meanwhile AI companies are like "but what if we went the other direction?"
"I have a PhD in hammerology and this is a nail"
I think rather than the entire OS being built by AI, this kind of research can evolve to make agents understand the OS better.
If it ever becomes a reality, we should call it the Bloat System (or BS for short). Its initialization process should be called the Bloat Loader, its memory, the B-RAM (when the bull meets the ram), its storage is the Slop Disk, and its programming language (prompts), the B#ъ (BShard) language, used to create Bь (BSoft) programs. Uninstalling it is called debloating your PC.
OSes behave the way they do because of the interaction of a large number of files, most of which are not visible on screen. To make a truly neural OS, one would have to train the model to represent the entire OS in its latent space. Not doing that leads to exactly what was shown in the video. There are older, hybrid ideas, such as AIOS, which is better because it doesn't try to replace the highly optimized architecture with an inefficient (in this case) approximation.
me blowing up the imaginary operating system by opening 6 terminal emulators and running dmesg on loop on every one of them