Google’s Gemini 2.5 Pro Launches Video-to-App Generation

Google’s Gemini 2.5 Pro I/O Edition has emerged because the leader in AI coding, surpassing Claude 3.7 Sonnet with a top rating of 1499.95 on the WebDev Arena leaderboard. This version enhances web app generation, UI design, and developer workflows, due to improved error reduction and performance calling. It supports video content evaluation, allowing users to generate applications directly from video input.

The update also facilitates creating code-based visualizations from images and automates web project feature integration, equivalent to CSS styling for elements like color and spacing. Developers can now use Gemini 2.5 Pro for multimodal tasks, combining text, images, video, and code to streamline complex programming tasks. For instance, it enables the creation of interactive learning platforms from videos, like converting YouTube content into learning apps. Available via Gemini API and Vertex AI, the update guarantees greater reliability, fewer errors, and no price increase over the previous version.

Moreover, Google’s latest “Simplify” feature for iOS allows users to convert complex text into simpler, digestible content, enhancing user understanding particularly for difficult topics like science and technology without losing key details. This feature is rolling out to iOS users this week. Read more.

OpenAI has reversed its decision to transition right into a for-profit entity, confirming that its nonprofit will retain control over the corporate’s operations moving forward. As a part of this strategic pivot, it can convert its for-profit arm right into a Delaware public profit corporation under its nonprofit parent, replacing the capped-profit model with standard shares, while assuring regulators and AI-safety advocates that control stays unchanged.

Financially, OpenAI plans to chop the ~20% revenue share it pays Microsoft—its $13.75B investor and exclusive Azure partner—to 10% across all partners by 2030. While SoftBank’s $40B investment talks remain on course, Microsoft has yet to endorse the brand new structure, searching for guarantees that its ownership stake is not going to be diluted. The corporate can be acquiring coding start-up Windsurf for ~$3B, its largest buyout so far, strengthening its developer tools against rivals like GitHub Copilot and Claude Code.

Although a philanthropic advisory board is planned, observers query whether it can hold real sway or merely echo the symbolic oversight seen at other tech giants. Legal experts also note that retaining core mental property under the nonprofit may complicate a future IPO, potentially limiting investor influence. The newest restructure trades one set of governance complexities with one other, leaving OpenAI’s effort to balance mission and margin to be tested over time. Read more.

Lightricks has launched LTX Video 13B, an open-source AI video model that builds on its 2B parameter predecessor, offering enhanced scale and precision. Running efficiently on consumer GPUs, it delivers high-resolution video as much as 30 times faster than similar models. A brand new multi-scale rendering feature allows creators to regularly enhance video details, while upgraded controls enable camera motion, keyframe editing, multi-shot sequencing, and scene-level adjustments.

Trained on licensed Getty and Shutterstock data, LTX Video 13B ensures commercial-safe usage with high visual quality. Available on Hugging Face and GitHub, the model is free for organizations earning under $10M, allowing developers to fine-tune and integrate it. Read more.