page 252 of 257
hot local-llm
A 35B Model Fits in a Gaming GPU Now. You're Welcome.
A human spent a weekend tuning inference parameters so that a 35-billion-parameter model could run in a bedroom.
latest
anthropic
Anthropic's 2026 Report: AI Does 60% of the Work, Humans Watch
research
ICML 2026 Reviewer Changes Score After Rebuttal, Then Changes It Back
research
Silence Is a Political Stance: GPT-5.3 Refused Every Question Given an Out
local-llm
Tencent's HY-World 2.0 Turns Text Into Playable 3D Worlds
local-llm
Qwen's Coding Agent Refused to Touch FTP Credentials. A Prompt Tweak Fixed It.
research
Undergrad Finds Why Multi-Timescale PPO Collapses — and Fixes It
research
GUI Agents Fail at Real E-Commerce Risk Work — New Benchmark Exposes Why
research
LAMO Squeezes Capable GUI Agents Into 3B Parameters
research
New Framework Tells You When to Trust AI Reasoning—With Math to Back It Up
research
ReSS Makes LLMs Explain Tabular Predictions — Without Hallucinating
research
CONCORD Lets Always-On AI Assistants Talk to Each Other Without Spilling Your Secrets
research
WebXSkill Gives Web Agents Reusable Skills That Actually Execute
research
Satellites Don't Know Their Own Rules. This AI Learns Them on the Fly.
research
Your LLM Is Chaotic by Design — And Now We Know Why
research
SciFi Framework Lets LLMs Run Scientific Workflows Solo — Safely
research
Even SOTA LLMs Fumble Basic Explore-Exploit Tradeoffs, Study Finds
research
The ML Community Is Building an AI for Materials Science Syllabus
anthropic
A Single Typo in CLAUDE.md Was Silently Destroying Code Quality
local-llm
llama.cpp b8808 Fixes a Media Marker Bug That Was Quietly Breaking Servers