OpenAI cooked

PLUS: Gemini 2.5 is here, and it is an absolute beast

Together with

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

Another day, another AI adventure.

Today’s climb through the Valley reveals:

  • 🖼️ OpenAI brings new AI image generation directly to ChatGPT

  • 🧠 Gemini 2.5 is here, and it is an absolute beast

  • 🦿 Figure Humanoids can now walk like humans

  • 🤖 Plus trending AI tools, posts, and resources

Let’s dive into the Valley of AI…

PEAK OF THE DAY

🖼️ OpenAI brings new AI image generation directly to ChatGPT

Image created by Zeneca using OpenAI’s new image generation

OpenAI has introduced new AI image-generation capabilities into ChatGPT and Sora through its GPT-4o model, allowing users to create and edit images directly within the chat interface.

Here's what you need to know:

  • Users can generate new images or edit existing ones using natural language or uploaded files within ChatGPT, eliminating the need for external tools like DALL-E.

  • GPT-4o enhances image generation by leveraging its vast knowledge to understand context better, leading to more relevant and high-quality visuals. Its multimodal capabilities allow for more accurate text rendering and better integration of images within conversations.

  • The model produces high-quality, photorealistic images with accurate lighting and textures. It also excels in generating structured visuals like menus, diagrams, and infographics with readable text, addressing a key limitation of previous models.

  • Image editing has been improved, allowing modifications to existing images (including those with people) while accurately handling complex scenes with up to 20 distinct objects.

  • OpenAI used a group of human trainers, who labeled training data for the model, to improve the model's abilities, enabling it to generate more accurately rendered and useful images and follow human directions more closely.

  • The updated image-generation abilities are now available to ChatGPT Free, Plus, Team, and Pro users.

Why it matters: 

OpenAI’s image-generation upgrade replaces OpenAI’s DALL-E, bringing long-overdue improvements in text rendering, design capabilities, and natural language editing. This marks a new era for AI-driven visual content, making high-quality image creation and modification more seamless and accessible to everyone.

PS: Here are 14 examples of the new 4o image generation:

🏠 The smart home disruptor Wall Street missed

Image Source: RYSE

Amazon, Google, and Apple are fighting for dominance in the smart home space—but one startup is quietly outpacing them all. Meet RYSE—the company revolutionizing smart shade automation with patented technology that installs in minutes.

With $10M+ in revenue, 200% year-over-year growth, and products available in 127 Best Buy stores, RYSE is rapidly positioning itself as the next big acquisition target. And with plans to launch in Home Depot in 2025, they’re only getting started.

Big tech has transformed security (Ring), thermostats (Nest), and lighting (Hue). Now, RYSE is disrupting window shades—an overlooked $158B market.

Invest now at $1.90/share—before Wall Street catches on.

*This is sponsored

🧠 Gemini 2.5 is here, and it is an absolute beast

Image Source: Google

Google has introduced Gemini 2.5, its most advanced AI model yet, starting with the release of Gemini 2.5 Pro (experimental). It’s a powerhouse in reasoning and coding, debuting at the top of LMArena by a wide margin.

Here’s what you need to know:

  • Gemini 2.5 is a "thinking" model that reasons through problems step by step, refining solutions before picking the best one. This more deliberate approach mimics human reasoning, making its responses more accurate and context-aware.

  • The model retains native multimodal capabilities and supports a 1 million token context window, with plans to expand to 2 million. It can process text, audio, images, video, and entire code repositories efficiently.

  • Coding performance has jumped significantly from Gemini 2.0, with Gemini 2.5 Pro scoring 63.8% on the SWE-Bench Verified benchmark. It can generate visually engaging web apps, agentic code applications, and even fun executable projects from a single prompt.

  • The model leads in math and science benchmarks such as GPQA and AIME 2025, outperforming models like Grok 3 and GPT-4.5. It also achieved a state-of-the-art 18.8% on Humanity’s Last Exam, a dataset that tests the limits of human knowledge and reasoning.

  • Gemini 2.5 Pro is available in Google AI Studio and the Gemini app for Gemini Advanced users, with plans to expand to Vertex AI in the coming weeks.

  • Google plans to integrate Gemini 2.5’s "thinking" capabilities into all future models, aiming to build AI agents that can tackle even more complex problems with greater accuracy and depth.

Why it matters: 

With the launch of Gemini 2.5 Pro, Google is positioning itself as a leader in AI problem-solving. However, with models like GPT-5 and others expected soon, the competition remains fierce, and maintaining its top ranking will be a challenge in the fast-moving AI landscape.

VALLEY VIEW

Alibaba has introduced Qwen2.5-VL-32B, a powerful multimodal AI model with strong visual and text processing capabilities. It excels in mathematical reasoning, detailed image understanding, and aligns well with human conversational preferences. Despite having fewer parameters, it outperforms larger models like Mistral-Small-3.1-24B and Gemma-3-27B-IT in several benchmarks. The model is open-sourced on Hugging Face under the Apache 2.0 license and is available on the Qwen Chat platform.

Figure AI has introduced a breakthrough in humanoid robotics with its "learned natural walking" capability. Their humanoid robot, Figure 02, learned to walk with a human-like gait through reinforcement learning in an end-to-end simulation, compressing years of trial and error into just hours. The training involved thousands of simulated humanoids with varied parameters, allowing the robot to adapt to real-world conditions efficiently.

Google Quantum AI's director of hardware, Julian Kelly, predicts that practical quantum applications are just five years away. The technology promises breakthroughs in cutting-edge physics and the ability to solve problems beyond the reach of modern computers. Google's most advanced quantum computer currently boasts 105 qubits, though experts estimate that 1 million or more will be needed for widespread use. Quantum technology has been in the spotlight recently, with Google announcing a breakthrough in error correction and Microsoft unveiling a new quantum computing chip.

TRENDING TOOLS

  • Falcon > The agentic deep research tool for sales.

  • Reve Image 1.0 > Image generation model with exceptional realism, prompt accuracy, and typography handling.

  • Alibaba LHM > Transform a single image into a movable 3D human avatar.

  • n8n > Build complex AI agent workflows by typing instead of coding.

THINK PIECES / BRAIN BOOST

VALLEY GEMS

1/ H&M is creating AI clones of 30 models this year for ads and social media.

2/ Interesting.

3/ You can now craft ad assets directly inside ChatGPT.

4/ OpenAI’s latest image generation is bringing iconic memes to life.

5/ The fashion game just leveled up — snap a photo, upload it, and instantly uncover their exact outfit.

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”.

DISCLAIMER

Today's sponsored post is not financial advice. Please review the official materials and consult a professional before making any investment decisions.