Microsoft released Quizzes for Copilot! Should be available on both web and mobile.

An upgraded "gemini-2.5-flash-native-audio-preview-09-2025" Gemini live model is now available on AI Studio!

Testing time 👀

Google AI Plus plan (budget option) expanded to 40 more countries!

Angola, Bangladesh, Benin, Bolivia, Burkina Faso, Cambodia, Cameroon, Côte d'Ivoire, Egypt, El Salvador, Ghana, Haiti, Honduras, Indonesia, Kenya, Kyrgyzstan, Laos, Mali, Mexico, Moldova, Morocco, Mozambique, Nepal, Nicaragua, Nigeria, Pakistan, Papua New Guinea, Philippines, Rwanda, Senegal, Tajikistan, Tanzania, Togo, Tunisia, Uganda, Ukraine, Uzbekistan, Vietnam, Yemen, Zambia and Zimbabwe.

Suno V5 is being rolled out to Pro users in Beta 🔥

Qwen Chat now has a Travel agent that can plan and organise your trip according to the given preferences.

Google released a new Mixboard experiment on @GoogleLabs in the US to let users play with a nano-banana-powered concepting board.

Qwen released the "Qwen3-LiveTranslate-Flash" for Multimodal Interpretation. The model speaks 10 languages, understands 18 and can also read lips and gestures.

Benchmarks 👀

Mistral AI tests Tone controls and custom styles in Le Chat

Mistral AI is developing new Le Chat features focused on customization, including a prompt cloud and a configurable Tone system for tailoring assistant behavior. These tools target both individual users and organizations seeking tailored outputs.

🗞 #mistral

Jules SWE agent from Google can now perform based on the PR feedback. No need to leave GitHub anymore.

“Jules-coding” 👀

BREAKING 🚨: OpenAI is testing new Agent upgrades for ChatGPT as "Agent with truncation" started appearing in the list of Alpha models. This is possibly a GPT-5-powered version of ChatGPT Agen.

Mistake rollout 👀

Copilot Portraits started appearing to some Pro users in US on Copilot Labs. These are some of the example portraits which users will be able to talk to.

h/t @LanceAdams

Microsoft adds Anthropic models to Microsoft 365 Copilot

Microsoft 365 Copilot now includes Claude Sonnet 4 and Claude Opus 4.1 from Anthropic, offering enterprise users AI model options tailored for complex reasoning, research, and automation tasks, available via the Frontier Program.

🗞 #claude

Claude released a possibility to move conversations to specific projects! It was a quite requested feature to my knowledge, good to see that Anthropic finally addressed it.

Move it 👀

OpenAI is testing a new Voice Mode UI on iOS in the latest update. Looks like it has new prompt suggestions! It is either an AB test, either a gradual rollout.

Did you get it? 👀

Google AI Studio will get an option to save secret env vars for Gemini Apps in the advanced settings.

OpenAI tests ChatGPT Agent upgrades powered by new Alpha models

OpenAI briefly surfaced experimental agent models in ChatGPT, suggesting ongoing development of tool-augmented workflows and new system prompt strategies. These early tests appear tied to preparations for an upcoming agent rollout.

🗞 #chatgpt

OpenAI prepares an adaptable "Daily Pulse" feature for ChatGPT

The ChatGPT iOS app update introduces a redesigned Voice Mode UI, Apple’s Liquid Glass support, and hints of a daily summary feature tied to user feedback. These additions suggest a shift toward a more adaptive, personalized experience.

🗞 #chatgpt

Kimi AI released "OK Computer", a new agent mode which can build complete websites.

"An agentic model with its own computer, K2 now has true agency."

Google prepares source search via Google Drive on NotebookLM

Google's NotebookLM is evolving to support research workflows, with updates enabling document sourcing from Drive, potential web-based deep research features, and possible audio/video overview tools for business and education use cases.

🗞 #notebooklm

Mistral AI is rolling out a possibility to index websites in Libraries. Now you can add them as sources to be grounded in Le Chat responses.

Let Le Chat do RTFM for you 👀

BREAKING 🚨: ChatGPT Pulse is rolling out to Pro users on mobile to deliver personalised reports which can be tailored to the user’s needs every day.

Pro testing time 👀