Comp-X: On Defining an Interactive Learned Image Compression Paradigm With Expert-driven LLM Agent
arxiv.org·1d
🧠Neural Compression
Show HN: A short story on developing a long-context World-Model with no money
francesco215.github.io·19h·
Discuss: Hacker News
🧠Learned Codecs
Deep Dive: OpenAI's GPT-OSS
dev.to·10h·
Discuss: DEV
📊Quantization
An inner-speech decoder reveals some mental privacy issues
arstechnica.com·2h
👂Psychoacoustics
FFmpeg 8.0 Released
ffmpeg.org·10h·
Discuss: Hacker News
🎬AV1 Encoding
I test audio kit for a living, and these earbuds' death-proof battery and regal design blew me away
techradar.com·4h
💿FLAC Archaeology
Show HN: Novel GPT-2 sampling and memory architecture
github.com·4h·
Discuss: Hacker News
💎Information Crystallography
My favorite Google AI features from the Pixel 10 launch
engadget.com·1d
🎙️Whisper
AI Training: What Creators Need To Know About Copyright, Tokens, And Data Winter
techdirt.com·14h
🤖AI Curation
Twittering Machines: LS1c is a magical looking glass
grimmaudio.com·22h
📼Audio Cassettes
A universal rhythm guides how we speak: Global analysis reveals 1.6-second 'intonation units'
phys.org·2d
🗣️CMU Pronouncing
Shotcut is a free and open-source video editor, available for all major platforms
techspot.com·17h
🎬Video Codecs
V-JEPA 2: Scaling V-JEPA
gonzoml.substack.com·22h·
Discuss: Substack
📊Learned Metrics
Show HN: Image Description Generator – auto-generate alt text from any image
image-description-generator.net·4h·
Discuss: Hacker News
🗜️LZW Variants
Unplug and Play Language Models: Decomposing Experts in Language Models at Inference Time
arxiv.org·1d
🧠Intelligence Compression
How I learn about generative AI
blog.pamelafox.org·4d·
Discuss: www.blogger.com
🤖Grammar Induction
Speeding Up AI Coding Assistants Using Deterministic Feedback
proxymock.io·21h·
Discuss: Hacker News
📼Tape Combinators
Cracking the Density Code: Why MAF Flows Where KDE Stalls
towardsdatascience.com·20h
🔗Tailscale
LLMs are NOT Turing Complete (at train time), we need "train time recurrence"
fchaubard.github.io·10h·
Discuss: Hacker News
📼Tape Combinators
Quantization Meets Spikes: Lossless Conversion in the First Timestep via Polarity Multi-Spike Mapping
arxiv.org·2d
🧠Neural Compression