AI Daily Briefing — 2026-06-25

The rapid expansion of artificial intelligence is hitting a critical bottleneck: access to real-time, structured web data. While AI models have advanced significantly, the internet itself was never built for automated data retrieval at scale. Much of the information needed to train and ground AI systems remains blocked, unstructured, or outdated. To overcome this, experts say a new infrastructure layer is emerging—one designed to help AI models navigate hundreds of millions of domains and billions of new URLs created each week.

“The data suggests there's far more data out there,” says Or Lenchner, CEO of Bright Data. “Think of the universe: It's out there, but you don't know what you don't know.” This infrastructure must handle millions of simultaneous interactions across websites varying by geography, language, and access rules. Without it, AI systems risk relying on stale snapshots, leading to poor decisions and eroded user trust.

Speed and freshness are now essential. Real-time data retrieval reduces hallucinations and grounds AI outputs in current, verifiable information. A recent survey found that 56% of AI practitioners believe businesses need access to live web data to improve trust in AI outputs. Yet many systems still struggle, even with retrieval-augmented generation (RAG). According to Gartner, 60% of AI projects lacking AI-ready data will be abandoned by year’s end.

“If it can't retrieve real-time information, it lacks context,” Lenchner says. “In a business setting, that's not acceptable anymore. Stale answers lead to bad decisions and disappointed consumers.” As organizations race to keep pace with dynamic markets, the ability to access fresh, structured data at scale may define the next wave of AI innovation.

Meta is rolling out a new standalone app for creators, transforming its existing Creator Studio tool into an AI-driven companion designed to help users grow their audiences on Facebook. The move is part of Meta's broader strategy to retain creators amid stiff competition from platforms like TikTok and YouTube, while also reducing reliance on third-party AI tools such as ChatGPT for content planning and performance analysis.

The app, currently in testing with select creators, integrates Meta's recently launched AI creator assistant. This assistant offers personalized recommendations based on a creator's content style, audience engagement, and goals. Instead of manually sifting through dashboards, creators can ask conversational questions like "When should I post?" or "What are people saying in my comments?" and receive quick, tailored answers.

Additional features include an AI-powered comment tool that surfaces key comments and drafts replies in the creator's tone, which can be edited and approved before posting. Each day, the app presents a feed of priorities, such as reviewing new post performance, tracking goal progress, and flagging comments needing responses. This launch follows a series of new Meta apps, including Forum for Facebook Groups and Instants for disappearing photos, as CEO Mark Zuckerberg has emphasized building more apps through AI-driven efficiencies.

Agility Robotics, the Oregon State University spinout known for its bipedal robot Digit, is going public through a merger with Churchill Capital Corp XI. The deal values the company at roughly $2.5 billion and is expected to generate over $620 million in proceeds, including $200 million from institutional investors. Agility plans to use the funds to ramp up production of its next-generation Digit v5, fulfill existing orders, and expand its customer base.

Digit is already deployed at nine customer sites, including Schaeffler, GXO, Toyota Motor Manufacturing Canada, and Mercado Libre. The company has secured more than $300 million in multi-year orders for the new model and has a pipeline of over 30 potential customers evaluating large-scale deployments. Agility has backing from Amazon, Nvidia, SoftBank Vision Fund 2, and DCVC.

CEO Peggy Johnson said humanoid robots will drive productivity and supply chain resilience, helping enterprises address labor shortages and integrate AI-powered automation. The combined company is expected to trade under the ticker AGLT on a North American stock exchange.

Figma has rolled out a significant update, introducing direct code layers, native animation support, and new AI-powered features. The design platform now allows teams to clone repositories and extract code-based flows directly onto its collaborative canvas, bridging the gap between design and development. Chief product officer Yuhki Yamashita emphasized that the goal is to enable rapid experimentation, allowing designers, product managers, and engineers to iterate freely without worrying about production-ready code.

In addition to code integration, Figma now supports animations, transitions, and 3D transforms natively, eliminating the need for external tools. Users can also apply shader effects and fills using AI. The company is further integrating its node-based tool, Weavy, acquired last year, to let designers run workflows across multiple AI models. Later this year, users will be able to generate Weavy workflows directly within Figma.

The update also enhances Figma's AI assistant, allowing users to create repeatable skills via text prompts and connect external tools like Notion, Excel, or GitHub for added context. A new feature enables users to build custom plug-ins—such as layout generators or vector path tracers—simply by describing them. These additions aim to streamline design workflows and reduce reliance on third-party software.

OpenAI has introduced its first custom-designed inference processor, developed in partnership with Broadcom. The chip, codenamed Jalapeño, is tailored specifically for running AI models in response to user commands. OpenAI says its own AI models helped design the processor, and early tests indicate it delivers significantly better performance-per-watt than current alternatives.

The move marks a strategic step for OpenAI to reduce reliance on Nvidia GPUs, following similar efforts by Google and Amazon to build custom AI accelerators. While tasks like pre-training will likely still depend on Nvidia hardware, the new chip aims to lower inference costs, which could improve the company’s bottom line. OpenAI president Greg Brockman noted that the company focused on underserved workloads to accelerate what’s possible.

Jalapeño is optimized for real-time coding models and low operating costs, according to OpenAI. The company emphasized that it is now designing infrastructure across the entire stack—from chip architecture to deployment systems—to make its models faster, more reliable, and more affordable. This vertical integration allows each layer to be fine-tuned toward the same goal.