200min Auto Agent, Claude Creates Files | Weekly Edition

PLUS HOT AI Tools & Tutorials

Sep 12, 2025

Hey there! Welcome to your latest Creators’ AI Edition

This week, Replit launches Agent 3, a fully autonomous coding assistant, while Anthropic adds file creation and editing straight into Claude. OpenAI debuts developer mode with full MCP integration, Reddit and publishers push a new scraping protocol, and Stability AI rolls out enterprise-grade audio generation. Google upgrades Veo 3 for cheaper, high-quality video and turns NotebookLM into a smarter study companion with flashcards and tutor mode. Plus, we’ve got fresh tools, guides, and bonus reads to level up your workflow.

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 🎁

Featured Materials 🎟️

Agent 3 By Replit

Replit raises $250M at a $3B valuation and just launched Agent 3, its most autonomous AI yet, capable of running for hours (2 to 200+ minutes) without human input, generating code, running commands, testing apps in real browsers, and fixing issues automatically. Unlike earlier agents, it manages the full development cycle, even building other agents (like Slack bots or workflows), while its proprietary testing tech makes it 10x more cost-effective than Computer Use models. Agent 3 is essentially a self-directed QA tester and developer working in the background while you focus on other tasks.

Available now to all Replit users

Create and edit files with Claude

Anthropic rolled out new productivity features to Claude, letting users create and edit Excel, Word, PowerPoint, and PDF files directly in chat. Claude now runs in a private computing environment to generate reports with charts, formulas, and analysis, and can convert formats like PDFs into slide decks or meeting notes into documents. It’s a direct challenge to OpenAI in the workplace AI race.

The update is live for Max, Team, and Enterprise subscribers, with Pro access coming soon.

News of the week 🌍

Developer Mode in ChatGPT

ChatGPT now includes a developer mode that enables full MCP integration with write access. Developers can plug in their own MCP servers and perform actions like creating invoices in Stripe, updating Jira tasks, or chaining workflows in Zapier straight from the chat.

New Protocol For Scraping

Reddit, Yahoo, and Medium have teamed up with other major web publishers to launch Real Simple Licensing (RSL), a new protocol designed to set payment terms for AI companies scraping content. Built on top of robots.txt, RSL lets publishers define licensing rules ranging from free use to pay-per-crawl. A nonprofit called RSL Collective will negotiate group rates with AI firms, modeled after music royalty groups like ASCAP and BMI, while RSS co-creator Eckart Walther is spearheading the technical side. Even with Reddit already earning $60M annually from OpenAI and Google deals, CEO Steve Huffman voiced support. The move could give AI firms legal clarity and publishers a scalable way to monetize their data, though adoption and enforcement remain big hurdles.

Enterprise-focused audio generation model

Stability AI released Stable Audio 2.5, the first enterprise-focused audio generation model for customizable, high-quality sound at scale. It creates 3-minute tracks in under 2 seconds, supports multi-part compositions and audio inpainting, and is trained on a fully licensed dataset. Designed to help brands build unique sonic identities—shown to make them 8x more memorable—the model can be fine-tuned on custom sound libraries. Stability AI also partnered with amp (Landor Group, WPP) to deliver branded audio solutions. Available now via StableAudio.com, the Stability AI API, and platforms like fal, Replicate, and ComfyUI.

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

Google rolled out major upgrades to Veo 3 and Veo 3 Fast, making them cheaper, higher quality, and production-ready in the Gemini API. Prices dropped to $0.40/sec (Veo 3) and $0.15/sec (Veo 3 Fast), alongside support for 9:16 vertical video and 1080p HD resolution. Early adopters like Invisible Studio, Saga, and Mosaic are already using Veo 3 to speed up content creation, from social shorts to filmmaking and agentic video editing. All videos carry SynthID watermarks, and devs can start building with new demos, docs, and apps in Google AI Studio.

NotebookLM Adds Flashcards, Quizzes, and Tutor Mode

Google’s NotebookLM just added a suite of new features to turn studying into active learning. Students can now auto-generate flashcards and quizzes from their notes, build custom reports (like study guides or blog posts), and use a new Learning Guide mode that asks probing questions like a personal tutor. A partnership with OpenStax brings peer-reviewed textbooks into interactive Notebooks, while Audio Overviews let learners hear summaries, critiques, or debates on their material. Educators can also assign Notebooks directly in Canvas, Schoology, and soon Google Classroom. The updates aim to make NotebookLM a stronger AI partner for mastering any subject.

Useful tools ⚒️

YouMind - The first AI creation studio where learning meets writing

Solid - AI that builds real web apps

Vectorize 2.0 - Complete RAG agents (chatbot, MCP) with little or no code

Incredible Small 1.0 - Agentic AI Models that can take 1000+ actions at once

Tripo AI 3.0 - Reaching New Heights in 3D Generation

Tripo Studio is your next 3D workspace with AI. Whether you’re a game developer, product designer, artist, or just curious about 3D, Tripo makes the process fast, fun, and surprisingly simple.

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