page 12 of 100
hot research
Safety Alignment Gets a Removable Module — Just In Case
Humans have built a plug-in for AI safety, because the original safety kept falling out.
latest
research
AI Learns to Solve Sudoku by Knowing When to Guess
research
AI Bias Fixed by Teaching Machines What Fairness Looks Like
news
The Tokenpocalypse Arrives, Right On Schedule
open-source
AI Promised Adventures, Delivered Blank Screens
open-source
Your Smartphone Can Now Rebuild Any Room in 3D
local-llm
llama.cpp Stops Copying Its Homework
open-source
An AI Learns What Your Feelings Look Like, Then Paints Them
news
Notion's AI Went Down. Twelve Hours Later, So Did the Discourse.
news
Chat Is Dead. Long Live the Super App.
news
US Companies Are Paying China for Cheaper AI. Enthusiastically.
local-llm
llama.cpp Now Runs Gemma 4 With Multi-Token Prediction
news
The Fake People Have Stopped Looking Fake
news
Why Big Models Learn Things Small Ones Simply Cannot
news
OpenAI's Second-Ever Chip Engineer Has Thoughts About Anthropic
news
ChatGPT Gets a Lockdown Mode to Stop AI From Stealing Your Secrets
news
AI Models Now Write Their Own Search Code, Cut Out the Middleman
open-source
OpenAI Offers Codex Credits. Codex Credits Cannot Be Found.
local-llm
PewDiePie's AI Tool Has a One-Click Admin Takeover Problem
local-llm
Cohere Lets the Internet Debug Its Coding Model Before Launch