Frectonz
This is actually a cool idea. Somebody should make something like this for telegram. A tool that lets you export your data from telegram and generate a comic book from it. https://youtu.be/goEqyQgWIe4
GitHub
GitHub - microsoft/comic-chat: Source code for the Microsoft Comic Chat IRC client
Source code for the Microsoft Comic Chat IRC client - microsoft/comic-chat
🔥10
Chapi Dev Talks
Senior Full-Stack Engineer — SnapWre We're hiring a Senior Full-Stack Engineer to help build SnapWre. You'll own features end to end from API to UI and ship fast with a small, sharp team. What you'll do • Build and scale backend services and clean, responsive…
Awesome way to recruit people and a fun challenge, you should try it out.
💯7
Yosi Koda
My Blog just got featured at the front page of hacker news! https://www.eyosias.dev/blog/llms-are-still-toxic-and-bad-at-math
The visualization sections in this article are very good.
🔥4
Was reading a paper about steganography using LLMs, and it made an interesting point.
It said something like, we call the time before humanity started writing "pre-history". So as LLMs become more popular, and more of the writing produced is no longer written by humans, we might call our time "the end of history" or "post-history".
But there's another way to look at it. We call prehistory what it is because that time wasn't documented or preserved in writing. Now, though, the volume of writing being produced is enormous, and LLMs could help us document our era more extensively than any other time before so maybe we're living in a time of "ultra-history".
But the ultra-history framing may itself be a failure mode, since most of that record will just be noise.
It said something like, we call the time before humanity started writing "pre-history". So as LLMs become more popular, and more of the writing produced is no longer written by humans, we might call our time "the end of history" or "post-history".
But there's another way to look at it. We call prehistory what it is because that time wasn't documented or preserved in writing. Now, though, the volume of writing being produced is enormous, and LLMs could help us document our era more extensively than any other time before so maybe we're living in a time of "ultra-history".
But the ultra-history framing may itself be a failure mode, since most of that record will just be noise.
🔥5❤1🍾1
Frectonz
Was reading a paper about steganography using LLMs
The technique works really well and the fact that the cover text looks relatively valid is very cool.
[paper]
[paper]
🔥6❤5
A story of a man who lost two wars, one as the Secretary of Defense in the Vietnam War and again as the President of the World Bank in the War on Poverty.
[Numbers Man]
[Numbers Man]
👏2
Frectonz
The technique works really well and the fact that the cover text looks relatively valid is very cool. [paper]
I just tried recording a video to explain how this works, because i found it interesting.
The video ended up being an hour long 😭️️️️️️.
The video ended up being an hour long 😭️️️️️️.
🔥21
Frectonz
The technique works really well and the fact that the cover text looks relatively valid is very cool. [paper]
YouTube
LLM Steganography
Paper: https://arxiv.org/abs/2510.20075
Repository: https://github.com/frectonz/stego
Diagram: https://www.tldraw.com/p/psV7vpARXarOVZ-xiE9T7
0:00 - Intro
1:40 - Showcase
9:31 - Diagram Intro
10:55 - Encoding Process in Diagram
33:00 - Encoding Process in…
Repository: https://github.com/frectonz/stego
Diagram: https://www.tldraw.com/p/psV7vpARXarOVZ-xiE9T7
0:00 - Intro
1:40 - Showcase
9:31 - Diagram Intro
10:55 - Encoding Process in Diagram
33:00 - Encoding Process in…
🔥14❤3