Howdy again. It’s Barsee, and welcome back to AI Valley.

Another day, another AI adventure.

Today’s climb through the Valley reveals:

• OpenAI launches new tools for building AI agents

• Google unveils Gemma 3

• OpenAI has trained an AI that's ‘really good' at writing

• Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

PEAK OF THE DAY

OpenAI launches new tools for building AI agents

OpenAI has unveiled suite of new tools designed to simplify the development of advanced AI agents, enabling developers to create systems capable of handling complex tasks autonomously.

Here’s the breakdown:

• At the heart of this launch is the new Responses API, which merges the simplicity of OpenAI’s Chat Completions API with the enhanced functionalities of their previous Assistants API. It enables developers to build custom AI agents that can manage complex workflows autonomously, reducing the need for multiple APIs.

• OpenAI has introduced three powerful built-in tools to enhance the capabilities and versatility of AI agents:

  1. Web search tool: This tool provides real-time information with citations, ensuring AI agents deliver accurate and up-to-date responses. Powered by the same models behind ChatGPT’s browsing feature (GPT-4o and GPT-4o-mini), it enables agents to access and reference the latest data seamlessly.

  2. File search tool: Designed for efficient information retrieval, this tool allows AI agents to sift through large document sets with ease. It introduces features like metadata filtering and a direct search endpoint, enabling faster and more targeted access to relevant content.

  3. Computer use tool: Leveraging OpenAI’s Operator model, this tool enables AI agents to visually interact with computer screens, performing tasks directly. Agents can navigate applications, complete forms, and interact with complex software interfaces, opening up new possibilities for automation.

• They have also introduced the Agents SDK, a toolkit designed to make it easier to manage one or more AI agents working together.

Why it matters: 

These tools make it easier to build AI agents. Developers can now create smarter systems with less effort. OpenAI also plans to add more features soon.

Online privacy is a boon for anyone who values their peace of mind

Have you ever considered how much of your data is “out there” while you’re traveling?

On just one trip, you may:

• Use your IP address and email to log in to hotel or airport WiFi

• Link your streaming accounts to an Airbnb TV

• Plug your phone into a compromised charging station

• Use your credit or debit card (a lot)

That’s a lot to leave to chance. Surfshark is a one-stop solution for keeping your digital self safe at home, abroad, and everywhere in between.

_{Thank you for supporting our sponsors!}

VALLEY VIEW

Google unveils Gemma 3, a powerful model that runs on a single GPU 

Google just launched Gemma 3, a new AI model with sizes from 1 billion to 27 billion parameters. It has a 128k token context window, so it can handle large data like high-res images and long texts. Gemma 3 is great at text and visual reasoning. Early tests show it beats bigger models like Llama-405B and DeepSeek-V3. You can access it on Google AI Studio and Hugging Face.

OpenAI says it has trained an AI that's ‘really good' at creative writing

OpenAI says its latest AI model is great at creative writing. Sam Altman shared a short story written by the model about "AI and grief" and said it was the first time AI-made writing moved him. He didn’t share when it will be released to the public, so it seems they’re still working on it.

Anthropic CEO: AI will handle 90% of coding soon

Dario Amodei, CEO of Anthropic, thinks AI will do 90% of coding in the next 3 to 6 months and maybe 100% within a year. Humans will still handle design and functionality, but tools like GitHub Copilot and Claude Code are already automating most of the work. AI is getting better at writing and fixing code from simple instructions.

A quarter of startups in YC’s current cohort have AI generated codebases

More startups are using AI to write code. In Y Combinator’s latest batch, about 25% of startups say AI generates up to 95% of their code. Tools like Cursor and GPT-4 Turbo help developers move faster by letting AI handle most of the coding while they review and refine it. This shift is changing the role of engineers, they focus more on product thinking and user experience, not just writing code.

NVIDIA introduces AI that turns images into realistic 3D videos 

NVIDIA has unveiled GEN3C, a generative video model that turns single or sparse-view images into photorealistic videos with precise camera control and 3D consistency. Users can define camera movements for dynamic and controlled perspectives, making it ideal for driving scenes, monocular videos, and immersive content. GEN3C can be used to enhance VR, AR, and filmmaking by simplifying the creation of realistic 3D environments and cinematic sequences.

Manus Al announced strategic partnership with Alibaba's Qwen Al team

Manus AI, a Chinese startup, is teaming up with Alibaba’s Qwen team to boost its AI agents. Manus will use Alibaba’s open-source models and computing power to handle more users. This partnership helps Manus grow and gives Alibaba a leg up over competitors like DeepSeek.

Wispr Flow for Windows: AI that frees you from keyboard & lets your thoughts flow naturally

Wispr Flow for Windows is finally here after winning over Mac users. Speak naturally and see your text appear 3× faster than typing, with AI that learns from your edits over time. It works in MS Word, Slack/Teams, emails, and even code editors—plus it supports 100+ languages, including UK English and Hinglish. Team plans make it ideal for businesses embracing voice-first workflows.

Get Started with Wispr Flow Today on MacOS or Windows

_{Thank you for supporting our sponsors!}

TRENDING TOOLS

1. Intercom > The human quality AI agent that talks with your customers 24/7, providing instant support across every channel. *

2. Eraser > Generate live diagrams from your codebase that stay up to date automatically.

3. Reworkd > Scrape 100s of unique websites using AI.

4. Perigon > AI with real-time context, automated signals.

5. IBM Granite 3.2 > A multi-modal AI model with great document understanding and complex problem-solving for business applications.

6. Cuckoo > Helps global teams chat effortlessly by instantly translating conversations in real-time.

VISUAL VALLEY

Soft Touch Images

Prompt: [Image Detail], --chaos 10 --ar 2:3 --sref 694725456 --profile tckem44 --sw 500 --stylize 500

THINK PIECES / BRAIN BOOST

1. Inside Google's investment $3 Billion in the AI start-up Anthropic.

2. What it's like to run an AI startup.

3. Cursor AI hits $100 million ARR and still haven't spent a dollar on marketing.

4. Is AI a bubble? It's complicated.

5. How will artificial intelligence affect wealth equality?

6. This new AI benchmark measures how much models lie.

7. How to learn anything 10x faster than anyone with AI.

VALLEY GEMS

1/ A glimpse into our future. Truckloads and truckloads of them in the future.

2/ A break down of why 4.5 matters and what it tells us about the future of AI.

3/ Testing Manus AI. Is it better than OpenAI Operator?

4/ Everyone will have IOS developer in their pocket.

SUNSET IN THE VALLEY

Thank you for reading today’s edition. That’s all for today’s issue.

💡 Help me get better and suggest new ideas at [email protected] or @heyBarsee

👍️ New reader? Subscribe here

Thanks for being here.

REACH 100K+ READERS

Acquire new customers and drive revenue by partnering with us

Sponsor AI Valley and reach over 100,000+ entrepreneurs, founders, software engineers, investors, etc.

If you’re interested in sponsoring us, email [email protected] with the subject “AI Valley Ads”.