Users can now upscale images generated by Nano Banana Pro to 2k and 4k on Google Flow. 4k option is only available to Ultra subscribers.

OpenAI opened App submissions for developers via its Platform. Builders can now submit Apps to appear in the upcoming Apps Directory on ChatGPT.

Google launches Gemini 3 Flash model with faster, low-cost results

Google introduces Gemini 3 Flash, a fast, cost-efficient AI model delivering strong reasoning and coding abilities. It surpasses Gemini 2.5 Pro on benchmarks, uses fewer tokens, and is now available across Google platforms for broad adoption.

🗞 #gemini

xAI launches Grok Voice Agent API for real-time voice apps

xAI has launched the Grok Voice Agent API, offering real-time voice capabilities and flat-rate pricing at $0.05/minute. It supports speech recognition, voice selection, companion personas, and integration with web and X data queries.

🗞 #grok

OpenAI opens ChatGPT to third-party apps after review process

OpenAI now allows developers to submit third-party apps to ChatGPT, following a review process. Approved apps become discoverable within the platform, marking a shift toward a native app ecosystem and a structured distribution model.

🗞 #chatgpt

Meta updates Edits app for creators with AI video effects and tools

Meta’s Edits app consolidates mobile video production with AI-powered tools, template support, multilingual text options, longer capture times, and direct Reels integration, offering creators a unified platform for content creation and export.

🗞 #meta

Google has released parallel execution for the Stitch design agent, allowing users to generate up to 5 different UI versions simultaneously.

Interesting take on how it may impact productivity, as parallelism is a good way forward.

Gemini x5 👀

OpenAI readies Codex Caribou model upgrade with GPT-5.2 base

OpenAI is preparing to release an updated Codex model, internally called Caribou, likely based on GPT-5.2. This version aims to raise baseline coding capabilities and may influence industry benchmarks and adoption among developers.

🗞 #chatgpt

Exclusive: Early look at Claude Tasks Mode Agent from Anthropic

Anthropic is testing a new task mode for Claude that supports step-by-step execution, multi-artifact output, and context selection. The interface mirrors NotebookLM but focuses on structured task planning and automation for advanced workflows.

🗞 #claude

BREAKING 🚨: Early look at Claude Task Mode agent workflow. What do we know so far?

- Claude Tasks will operate various Skills and MCPs to achieve the goal.
- Claude will ask clarifying questions before the execution or skip it after a timeout.
- Claude will generate an action plan for complex tasks.
- Users will be able to modify the plan on the fly or send additional context to the workflow while Claude is working.
- Generated artefacts can be previewed in a separate panel.
- All artefacts, Skills, noscripts and generated files will be stored in the working files directory.
- And more details in the post 👀

NotebookLM on Skills 🔥

xAI works on Imagine API and Grok API Playground

xAI has launched its Voice Agent API and previewed tools like Playground and Imagine API to expand Grok's capabilities. These updates support developers and signal a broader move to compete in the AI infrastructure market.

🗞 #grok

Google is hiding something from us

👀👀👀

Mistral released Mistral OCR 3, which outperforms existing enterprise document processing solutions as well as AI-native OCR systems.

BREAKING 🚨: Luma Labs is launching Ray 3 Modify, a new video generation model that allows users to modify existing videos by providing character reference images.

You can inject yourself everywhere now 👀

BREAKING 🚨: OpenAI is releasing GPT-5.2-Codex "Caribou" model today! An earlier code was updated with a new model name just recently.

Google is likely about to release Gemma 4 today, as "Google's Gemma models family" collection got updates just recently.

F5F5F5 👀

Not Gemma 4 but functiongemma-270m-it 👀

"FunctionGemma is intended to be fine-tuned for your specific function-calling task, including multi-turn use cases."