DeepSeek R1 Pure RL: 97% Cost Cut Challenges AI Training
DeepSeek R1 matches OpenAI o1 using pure RL at $294K training cost—27x cheaper API, no supervised fine-tuning needed. MIT licensed, benchmarks validated.
MCP Joins Linux Foundation: AI Rivals Unite on Open Standard
Anthropic donated Model Context Protocol (MCP) to the Linux Foundation’s new Agentic AI Foundation in ...
DeepSeek MHC Hits 16% Better AI With 7% Overhead
Chinese AI startup DeepSeek kicked off 2026 on January 1 with a technical paper proposing ...
Korea’s AI Sovereignty Fails: Naver Used Chinese Model
South Korea’s $1 billion national AI project just hit its first major test—and it’s failing. ...
Chrome Brings Back JPEG XL After 2022 “Obsolete” Kill
Google reversed its 2022 JPEG XL removal after Safari shipped it and developers built a ...
Harmattan AI Hits $1.4B Valuation: Defense Unicorn in 20 Months
French defense AI startup Harmattan AI closed a $200 million Series B led by Dassault Aviation, valuing the company at $1.4 billion and ...
Apple Picks Gemini After Siri 33% Failure Rate Forces $1B Bet
Apple and Google announced a multi-year partnership on Sunday, January 12, 2026, making Google Gemini ...
AI Coding Productivity Paradox: Why Devs Are 19% Slower
A September 2025 study by METR found developers using AI coding tools were 19% slower ...
Claude for Healthcare: Anthropic Responds 5 Days After ChatGPT Health
Anthropic launched Claude for Healthcare on Sunday, January 12, 2026, just five days after OpenAI ...
TimeCapsuleLLM: Student’s 1800s-Trained AI Accidentally Discovers Real 1834 History
A college student trained an AI exclusively on 1800-1875 London texts to study bias reduction. ...