OpenAI launches GPT-5.3-Codex for software tasks on paid plans

OpenAI's GPT-5.3-Codex targets end-to-end software tasks beyond code generation, showing improved performance on benchmarks and lower token usage. It is faster, supports cybersecurity use, and is available via paid ChatGPT platforms.

🗞 #chatgpt

GitHub lets Copilot subscribers run multiple AI coding agents

GitHub's Agent HQ update allows Copilot Pro+ and Enterprise users to run multiple AI coding agents within GitHub platforms, supporting task-specific agent selection, audit logging, and centralized workflow collaboration.

🗞 #github

Comet browser agent and Model Council features have been upgraded to Claude Opus 4.6 for Perplexity Max subscribers.

Claude Opus 4.6 claimed the first spot on Arena across text, code and expert categories.

Opus 4.6 scored 1496 on text 👀

ai-dot-com domain has been reportedly purchased for $70M by crypto-dot-com.

ai-dot-com will be launched during the Super Bowl as a decentralised platform for AI agents that can perform actions on behalf of users.

During the launch phase, users will be able to claim their usernames.

"ai-com let's you create an agent with its own computer, so it can use any application and do any task that you might do. it's not a chat bot, but a fully functioning digital being."

Sora now allows generating videos and characters from images of people with stricter moderation.

"Starting today, eligible users can upload images with people to make videos in Sora, after attesting that they have consent from people featured and rights to upload the media."

Codex app for Windows is in the works 👀

h/t x@ajambrosino

ICYMI: Arena launches Max router to boost AI prompt accuracy

Arena has launched Max, a model router that uses over five million community votes to select the best language model for each prompt. Max leads in key benchmarks and supports reliable outputs in coding, math, creative writing, and expert tasks.

🗞 #ai

Exclusive: A new Gemini 3 Pro checkpoint spotted in A/B testing

Google is testing a new Gemini checkpoint in AI Studio showing strong performance in UI generation and SVG rendering. Though release details are unclear, the model draws developer interest and may influence upcoming AI product capabilities.

🗞 #aistudio

Google tests Personal Intelligence for NotebookLM conversations

Google is testing a “Personal intelligence” layer in NotebookLM, allowing it to learn from chat history.

🗞 #notebooklm