Krea AI released an identity preservation feature for Krea 2 along with updated models and ComfyUI nodes. The update lets users maintain character appearance across multiple generations without manual reference work. The change targets creators who previously needed repeated prompt tweaks or external tools to keep faces and outfits consistent. Krea AI posted the announcement on X on July 8, 2026. The post states that the new models and ComfyUI nodes are already available. No

Microsoft and Amazon anchor nextgen nuclear deals as data center operators seek reliable power, while fusion startups target commercial delivery after 2030.

XR glasses reach daily use in 2026, with Meta and Apple hardware driving always-on visual capture that creates tension between seamless integration and user privacy controls.

Google expanded Gemini API Managed Agents with background execution, remote MCP server connections, custom function support, and credential refresh options. Developers can now run tasks asynchronously, connect directly to external model context protocol servers, and mix local business logic with built-in sandbox tools without extra middleware layers. The update targets production reliability. Teams previously had to build custom proxies or handle long-running tasks through po

Neuromorphic chips that mimic neurons are cutting AI energy use by up to 90 percent, but adoption requires new analog tooling and hybrid workflows that challenge mid-tier operators most.

Meta's CTO Andrew Bosworth acknowledges record-low employee morale amid ongoing restructuring and AI investments, highlighting tensions between efficiency goals and workforce strain.

A German regional court has ruled that Google bears legal responsibility for inaccurate AI Overview summaries, creating new compliance obligations for AI-powered search features.

Fun-ASR-Realtime was released by Tongyi Lab as a single real-time speech recognition model. It supports 30 languages and 16 dialects with reported semantic accuracy of 87.8 percent on internal industrial tests. The release focuses on East and Southeast Asian speech varieties. The model adds context understanding and dynamic hotword injection to resolve homophones and brand names during recognition. Streaming first-token latency stays in the low hundreds of milliseconds while

NVIDIA Kyber NVL144 delayed by over 12 months to 2028 Three months after Jensen Huang presented Kyber NVL144 at GTC, the project was pushed back more than twelve months. The delay moves full availability to 2028. At the same time, plans for an NVL72x2 back-to-back rack design were canceled, tightening the scaling options for Rubin Ultra systems. Announcement meets immediate revision NVIDIA showed the Kyber NVL144 configuration during the GTC keynote as part of its next-genera

Zuckerberg announced plans to build the Prometheus cluster, Meta's first single AI training site above one gigawatt of power capacity. The project forms part of a larger push to pull elite researchers, massive capital, and dedicated power infrastructure under one roof. The move comes as several labs race to secure the compute needed for next-generation models. The announcement was shared via X by Rohan Paul (source). Meta already operates large GPU fleets. Scaling one site to

Auto-translate features in browsers bypass cookie consent and degrade ad tracking precision, hitting independent publishers hardest.

JoyAI App released a user-generated content tool that creates custom digital humans from a single photo upload. According to JoyAI’s official announcement, the feature lets users “upload a single photo to create a personal virtual companion with matching voice” and delivers “industry-leading full-duplex conversation” that supports interruption at any time. The tool reuses the “Universal Doctor” technical base and integrates its language, voice and digital human models. JoyAI

Shengshu Technology released Vidu S1 at the 2026 Global Digital Economy Conference. The model supports real-time video calls and voice-controlled generation that can run for unlimited duration. The company built Vidu S1 on an autoregressive diffusion approach. It takes frames already created and voice instructions as input, then predicts what comes next frame by frame. One still image is enough to generate a character and assign a custom voice. No separate 3D modeling step is

WebKit released Safari Technology Preview 247 with a built-in MCP server. The server lets any Model Context Protocol client connect to a Safari window and pull DOM trees, network requests, console output, and screenshots. Developers no longer need to copy data between browser tabs and external tools. The integration runs locally after one setting change. What the update actually ships The MCP server registers four core tools in every Safari instance. browser_console_messages

Fable now handles more of its own decision making after Simon Willison tested advice from the Claude Code team. The change replaces fixed rules with open prompts that let the model judge when to add tests or call other models. Willison wrote about the experiments on his blog after a session at AIE. The core shift is simple. Instead of telling Fable exactly when to write tests, he now asks it to decide for itself. The same approach applies to other routine steps. The second ch

Wan Video added a music companion dance tool. Users upload one character and one song. The system then builds a video where the character moves in time with the beat. The feature supports five dance styles. These include street dance, tap dance, Latin dance, K-Pop, and Chinese classical dance. The tool matches steps to music beats without manual keyframing. How the Feature Works in Practice Alibaba Cloud posted the update on X. The post described a simple workflow. Load a cha

China released a draft update to its Internet Information Services Management Measures on July 3. The new version adds a full chapter titled "intelligent information services" that sets rules for AI providers. The draft requires companies to publish basic technical principles, list sources of training data, and mark any generated or synthetic content. It also bans forcing users into AI services or using algorithms to shape online opinion. Feedback closes on August 2. The full

Developers experimenting with Fable and other coding agents are converging on a simple operating rule: do not hand the model the whole repository and walk away. The short-leash method keeps planning, diff review, and merge approval in human hands while still using the model to accelerate execution. The approach pushes back on so-called YOLO mode, where an agent receives a broad objective and is allowed to edit freely for long stretches. In the short-leash version, developers

Kuaishou announced that 21 initial investors will put 2.028 billion dollars into Beijing Keling. The unit holds Keling AI assets. The move sets the post-money valuation at 18 billion dollars. HKEX announcement The filing also shows 15 more investors adding up to 766 million dollars. The total raise cap reaches 3 billion dollars. That amount equals 16.67 percent of the enlarged registered capital. Kuaishou expects Keling AI to start a Hong Kong listing process inside the next

Michael Butler, 44, drove a Tesla Model 3 into a Texas home and killed 76-year-old Martha Avila. Police records show his phone contained multiple searches for terms such as FSD too conservative and Tesla FSD overly cautious in the weeks before the crash, according to documents filed in the Texas court. Vehicle data later revealed the accelerator stayed pressed at 100 percent for six seconds with no brake input, pushing the car to 117 km per hour, as detailed in the NTSB preli