First screen, captured . Open the video.
Captured . Click for the whole screenshot.
101 points by handfuloflight | 10 comments
arXiv paper introducing Memorilla, a system that compresses document collections into query-dependent memory tokens in the input embedding space of frozen LLMs.
3 points by matt_d | 2 comments
arXiv preprint introducing Soft-MTP, a one-line change to multi-token prediction training that keeps only a KL self-distillation gradient signal.
171 points by theanonymousone | 39 comments
arXiv paper introducing Next-Concept Modeling, a coarse-to-fine text generation method using a multiscale residual VQ-VAE and parallel transformer decoding, faster than diffusion LMs.
15 points by christ_dev | 2 comments
GitHub repository for Reflota, an open-source, read-only Windows tool for recovering deleted and lost files from NTFS, FAT and exFAT disks, with phased disk imaging.
7 points by AmSach | 3 comments
GitHub repository for MAPLE, a Mixture-of-Experts pruning framework that compresses Gemma 4 26B-A4B to a 4.75 GB GGUF to run fully in VRAM on 6 GB GPUs.
660 points by gf74918 | 139 comments
GitHub README documenting how consumer Gemini with Connected Apps can pull Drive and Gmail content into chats that are subject to human reviewer annotation.
333 points by yurivish | 300 comments
Ajeya Cotra's Substack post arguing a METR report shows AI is already roughly doubling AI capabilities progress at Anthropic, implying rapid compounding toward superintelligence.
36 points by artninja1988 | 26 comments
Anthropic Frontier Red Team post reporting experiments suggesting small AI models may soon self-replicate on commercial IoT devices without datacenter access.
319 points by realsarm | 149 comments
GitHub repository for FastUI by Pydantic, a library for building web user interfaces declaratively in Python with React rendering.
1091 points by minimaxir | 695 comments
OpenAI announcement post previewing GPT-5.6 Sol, a next-generation language model, with a linked system card.
13 points by gmays | 3 comments
arXiv paper introducing UniEvo-VL, an on-policy self-distillation framework where a multimodal model uses its own self-critiques to improve image generation.
411 points by tplrbv | 72 comments
A long blog post analyzing how a small ~10M parameter transformer trained from Karpathy's tutorial turns attention outputs into next-token predictions.