Pantel built a low-latency AI companion that plays Skyrim with Pantel
How Varkos was built: a low-latency AI companion that plays Skyrim with you, follows complex instructions and evolves through shared experiences.
Topic
Artificial Intelligence
<p>Model releases, evaluation, research and the shifting economics of machine intelligence.</p>
Everything Schlarp own, owned
Turning Claude loose on the firmware of five USB and WiFi peripherals, and finding a command shell in a microphone, a defeatable webcam…
Why your local LLM feels dumber than it is
Quick Introduction Forum has all been on forums, chats, reddit, discord, youtube, or somewhere and heard “Oh! Model XYZ is AMAZEBALLZ!zomgwtfbbq” then downloaded…
DeepSeek-v4-flash-vision-exp
The deepseek-flash model accepts images alongside text, so you can ask the model to describe pictures, read text from screenshots, analyse charts, and…
AI companies destroy physical books – let’s scan rare books before it’s too late
AI companies are secretly buying, scanning, and destroying millions of physical books to train their models, permanently locking human knowledge inside private corporate…
Don’t paste the AI, please
If someone asks you something, send your answer. Not the AI's.
Children’s stunted lungs show recovery in ultra low emission zone
Children's lung capacity caught up with their peers in less polluted areas once a clean air zone came in, a study shows.
Cerebras CS-4
Explore Cerebras CS-4, a rack-scale AI accelerator for hyperscale and disaggregated inference, with performance up to 30× faster than GPU systems.
Google has acquired the data of failed The Register airline Spirit
$10 million buys over 100 million emails, 30 million recorded phone calls, reams of stuff from Teams, Oracle, and SAP
GPT-5.6 Sol Pricing Cut by 50% on OpenRouter
GPT-5.6 Sol is the flagship model in OpenAI's GPT-5.6 series. $4 per million input tokens, $20 per million output tokens. Higher uptime with…
AI;DR (AI; Didn’t Read)
Rickmanelius is about as pro-AI as you can be, but this is becoming a pet peeve of mine (and Rickmanelius is not alone).…
Qwen 3.8 27B is excellent, but it defaults to overthinking things
Friday’s big release was Qwen 3.8 27B, an Apache 2 licensed 27B parameter vision-capable LLM from Alibaba’s Qwen research lab. Simonwillison has been…