AGENTS.md, QWEN's 20B Open-Source Image | Weekly Edition

PLUS HOT AI Tools & Tutorials

Aug 22, 2025

Hey there! Welcome to your latest Creators’ AI Edition

This week, Alibaba drops a 20B open-source image editor, a new AGENTS.md standard promises to unify coding agents, and Google Photos gets conversational AI editing. Nvidia’s compact Nemotron models show big reasoning gains, OpenAI launches ChatGPT Go in India, and Gemini Live learns to guide you with visuals. NASA and IBM forecast solar flares with Surya, and ElevenLabs expands its agents with text-only chat. As always, we’ve got the latest tools, guides, and bonus reads to keep your AI game sharp.

Let’s dive in!

Featured Materials 🎟️
News of the week 🌍
Useful tools ⚒️
Weekly Guides 📕
AI Meme of the Week 🤡
AI Tweet of the Week 🐦
(Bonus) Materials 🎁
Keep your mailbox updated with key knowledge & news from the AI industry

Featured Materials 🎟️

A 20B Open-Source Image Editor

Alibaba’s Qwen team has launched Qwen-Image-Edit, a 20B parameter model for advanced image editing. It supports semantic edits like rotation, style transfer, and IP creation, as well as appearance edits that keep most of the image unchanged.

The model also enables bilingual text editing in Chinese and English, preserving fonts and formatting. Benchmarks show it outperforms rivals like Seedream, GPT Image, and FLUX.

With features such as object insertion, reflection handling, and step-by-step corrections, Qwen-Image-Edit brings state-of-the-art precision and aims to make professional visual editing accessible to everyone.

New AI Agent Standard

Every AI coding agent has had its own way of reading project instructions. Claude Code requires a CLAUDE.md, Google Jules looks for AGENTS.md, and the list keeps growing as more agents appear. This fragmentation has forced developers to maintain multiple files that essentially contain the same details. A new industry-wide initiative by OpenAI, Google, Cursor, AmpCode, FactoryAI, and Roo Code aims to solve this with a single standard: AGENTS.md. The file captures everything an agent needs: build commands, testing steps, security considerations, coding style, and deployment instructions, while supporting nested configurations for large and complex monorepos. With more than 20,000 open-source projects already adopting it, AGENTS.md establishes true interoperability so that developers can define their project context once and have it work across all major coding agents.

Google Photos Adds Conversational AI Editing

Google is rolling out a new conversational editing feature in Google Photos, starting with the Pixel 10 in the U.S.. Users can now simply ask Photos, by text or voice, to perform edits like “remove the cars in the background” or “restore this old photo”, without selecting tools or sliders. The editor also supports multiple requests in a single prompt, from quick fixes to creative edits like background swaps or adding objects.

To improve transparency around AI edits, Google Photos is introducing C2PA Content Credentials, showing users how an image was captured or modified. The feature will expand beyond Pixel 10 to Android and iOS devices in the coming weeks. Powered by Gemini models, Photos continues to evolve into a creative, AI-driven editing hub.

News of the week 🌍

NVIDIA’s New Model

Nvidia has unveiled Nemotron Nano 2, a new line of compact reasoning models ranging from 9B to 12B parameters. Despite their small size, the models deliver state-of-the-art reasoning performance while running at 6x the speed of comparable systems. This release highlights Nvidia’s push to make advanced reasoning more efficient and widely deployable on smaller compute footprints.

ChatGPT Go

OpenAI has introduced ChatGPT Go in India, a new subscription tier priced at Rs. 399 (under $5/month). The plan offers 10x higher message limits, 10x more image generations, 10x more file uploads, and 2x longer memory compared with the free version, while allowing payment in local currency. This marks a more affordable entry point for Indian users to access advanced ChatGPT features.

Gemini Live Gets Visual Guidance

Google is upgrading Gemini Live with new on-screen visual guidance, deeper Google app integrations (Calendar, Keep, Tasks, and soon Messages, Phone, Clock, Maps), and a more expressive audio model. Launching with the Pixel 10 on Aug 28, Gemini can now highlight objects via your camera, manage tasks across apps, and respond with more natural intonation, rhythm, and pitch, making it feel closer than ever to a true everyday AI assistant.

Thanks for reading Creators' AI! This post is public so feel free to share it.

NASA and IBM have unveiled Surya, a new AI foundation model trained on nine years of solar data to forecast solar flares and space weather with unprecedented accuracy. Using millions of images from NASA’s Solar Dynamics Observatory, Surya improved prediction accuracy by 16% over current methods, tracking phenomena like sunspots and solar wind speeds that can threaten satellites, power grids, and astronauts. By analyzing multiple wavelengths at once, it detects subtle solar shifts that humans often miss. Importantly, Surya has been open-sourced on HuggingFace, allowing researchers worldwide to leverage its solar forecasting abilities. With rising risks from solar storms amid growing space activity, Surya could be a critical shield against billions in potential damage.

ElevenLabs Agents Add Chat Mode for Text-Only Conversations

ElevenLabs has introduced Chat Mode, letting enterprises build text-only conversational agents as part of its Conversational Agent platform. Designed for cases where typing works better than speaking, like entering order IDs, handling simple issues, or when customers prefer chat, the feature can be deployed in minutes via SDK, API, or even a single line of HTML. Existing voice agents can switch to Chat Mode instantly, enabling seamless support across voice, chat, or both.

Share this post with friends, especially those interested in AI!

Useful tools ⚒️

Macaly 2.0 - AI website builder with built-in database, hosting & more

Ponder - AI-Powered Journal for Self-Reflection

Moises AI Studio - Introducing the first instrument-based AI music model

Stormy - AI agent for influencer marketing

Broxi AI - No-code AI agent builder. From text to AI agents in minutes

Describe what you want, and let Broxi Autopilot automatically build your AI Agents in minutes. Publish it anywhere using our APIs and automate thousands of tasks. From customer support automation to sales enablement, internal ops, etc, Broxi gets you covered.