DeepMind Cancels ‘Project Genie’: Instant Text-to-3D Navigation

Google DeepMind has launched Project Genie for US subscribers at $250 per month for AI Ultra. This innovative tool creates interactive 3D environments in real time—just input a prompt, and it generates a navigable scene at 1280×720 resolution and 24fps. Each session is limited to 60 seconds, allowing users to sketch, explore, and modify the generated output. With this initiative, DeepMind aims to train AI agents in simulated spaces rather than static datasets.

Gemini 3 Flash introduces Agentic Vision. Instead of merely recognizing images in a single shot, the AI can now examine, zoom in on, measure, and annotate objects step by step using Python. This results in a 5–10% increase in accuracy on vision benchmarks. Applications include reviewing architectural plans, labeling items with bounding boxes, and generating visuals from photographs, available via AI Studio and Vertex AI.

Google Maps now enables hands-free voice queries for pedestrians and cyclists without interrupting navigation. Chrome’s Auto Browse facilitates multi-step web tasks like trip planning, form filling, appointment scheduling, and subscription management. It leverages the Universal Commerce Protocol (developed with Shopify, Target, Etsy, and Wayfair) to search for products and apply discounts. Payment processing and social media postings pause for user approval. These features are only accessible to AI Pro and Ultra subscribers.

For developers, Stitch provides design systems, Figma export capabilities, and a Deep Design mode. Additionally, Jules, the coding assistant, now automatically addresses CI errors and integrates with Linear, Supabase, Neon, and Stitch. Future iterations will operate round-the-clock, managing backlogs and bug fixes autonomously. AI Pro/Ultra plans now provide Developer Program access and Cloud credits for AI Studio, Gemini CLI, and Vertex AI. Explore further.

OpenAI is prepping for a Q4 IPO and swiftly streamlining its product offerings while refining its long-term direction. The firm is engaging with Wall Street banks and has bolstered its financial team with two significant appointments: Ajmere Dale as Chief Accounting Officer and Cynthia Gaylor heading corporate finance and investor relations. This IPO is anticipated to generate substantial funds for computational infrastructure and AGI research, all while establishing a market valuation ahead of competitors like Anthropic.

OpenAI is retiring several outdated models next month, including GPT-4o, GPT-4.1, GPT-4.1 mini, and o4-mini. While GPT-4o still has a dedicated following for its friendly, conversational tone, only 0.1% of users engage with it daily. Most have transitioned to newer iterations (especially GPT-5.2). OpenAI is directing its resources toward high-engagement models that showcase improved personality, creativity, and reasoning. Access to retired models via the Developer API will persist.

Following a six-month stint, ChatGPT Agent is being discontinued. Initially launched with 4 million weekly active paying users, it quickly lost 75% of them. Many users found the offering unclear, as its virtual browser overlapped significantly with functions already present in other ChatGPT modes (coding, web research, image analysis). OpenAI is now moving towards more streamlined, well-defined specialized agents, beginning with tools like Shopping Research that effectively address single tasks.

Sora launched with 100,000 installs on its first day but has recently struggled to maintain momentum. Downloads plummeted by 32% in December and 45% in January. The rise of competitive platforms like Google Gemini for video generation and Meta’s Vibes, along with stricter copyright regulations, has hindered its growth. A limited licensing agreement with Disney has not substantially increased user engagement.

Elon Musk’s ventures are becoming increasingly interconnected. xAI has unveiled the Grok Imagine API, capable of converting text or single images into 15-second 720p videos featuring realistic motion, object interactions, continuity, and integrated audio and dialogue.

Users have the ability to edit scenes, add or remove objects, animate characters, and adjust lighting, seasons, or aspect ratios.

Grok ranks first on benchmark tests for both text-to-video and image-to-video, outperforming competitors such as Runway, Kling, and Veo 3.1, while being faster and more economical than Sora or Veo.

𝕏 is currently exploring the implementation of “manipulated media” labels to highlight edited or AI-generated visuals, although detection methods and appeals processes have not yet been fully defined. Tesla is gearing up for the production of Optimus Gen 3 set for production in Q1 2026, enhancing features like hands and establishing a dedicated assembly line in Fremont with a target of producing 1 million units annually for industrial, home, and medical uses.

Tesla has allocated $2 billion to xAI, integrating Grok into its vehicles and sharing energy systems. Initial merger discussions are underway about potential amalgamations of SpaceX, Tesla, and xAI, possibly before a SpaceX IPO. Musk has suggested that AI data centers in orbit could reduce operational costs through continuous solar energy and diminished cooling demands, supporting extensive AI functionalities on Earth and beyond. Learn more.

• Zuckerberg hints at agentic commerce tools and a major AI rollout in 2026

• Apple’s second-largest acquisition ever involves an AI firm focused on ‘silent speech’

• Anthropic releases metrics to monitor AI-augmented developer productivity

• Worldcoin WLD token surges as OpenAI considers a biometric social network

• Qwen3-Max-Thinking advances AI reasoning by reflecting on its thought process

• Zuckerberg’s vision for an AI-enhanced workforce

• Nvidia, Microsoft, and Amazon discuss a potential $60 billion investment in OpenAI’s upcoming funding round

• Satya Nadella affirms heavy usage of Microsoft’s Copilot AI

• ARC prize introduces open-source ARC-AGI-3 developer toolkit for AI assessment

• Amazon quietly identifies significant child exploitation content embedded in its AI training datasets

• InVideo and Anthropic launch AI-generated motion graphics

• Suno launches the “Sample” tool to transform sounds into music

• Cursor reduces codebase indexing time from four hours to just 21 seconds

• Robots now have the ability to learn new skills, similar to apps on your smartphone—thanks to OpenMind’s app store

• One of the largest time crystals has been created using IBM’s quantum heron chip

• Watch: Pentagon’s drone swarm test showcases one operator hitting three targets simultaneously

• World’s first exascale supercomputer accelerates plant research through AI integration

• Watch: Ai-Da makes history as the first humanoid robot to present a building design

• It’s a new era for gas thanks to data centers

• PwC broadens its partnership with Google Cloud with a $400M initiative in AI-powered security operations

• UMG files a $3 billion lawsuit against Anthropic for alleged copyright infringement of over 20,000 songs

• A groundbreaking flexible AI chip designed for smart wearables is thinner than a human hair

• LingBot-VLA introduces depth-aware cross-robot manipulation for real-world applications with minimal training

• AI models benefit from simulated internal debates, which significantly enhance their accuracy

• Apple updates its Pro suite, transforming it into a $12.99-a-month playground for the AI generation

• OpenAI accounts for nearly half of Microsoft’s future contracts, causing investor concerns

• Nvidia’s proprietary technology has enhanced DeepSeek’s AI models for China’s military

• While GPT-4 excels, human intelligence remains the ultimate benchmark

• AI may design drugs, but the FDA insists only humans can sign off

• ServiceNow plays both sides of the AI debate by naming Anthropic as its preferred partner right after an OpenAI deal

• The nation’s leading cyber defense official inadvertently uploaded sensitive government information to ChatGPT

• China allows ByteDance and Alibaba to purchase Nvidia’s H200 chips

• The Doomsday Clock reaches an alarming 85 seconds, highlighting unregulated AI’s role in global threats

• LM Studio 0.4.0 adds server deployment, REST API support, parallel requests, and an updated user interface

• Dubai embarks on building the first residential villa constructed by robots

• Samsung’s inaugural chief design officer aims to-transform AI into an experience rather than a mere tool

• Astronomers utilized AI to identify 1,400 ‘anomalous objects’ from Hubble Space Telescope data archives

• LinkedIn shifts away from traditional CVs, introducing a verification system to showcase actual AI competencies

• Silicon Valley bets $73 million on “digital concrete,” hoping to revolutionize traditional document formats

• Your next cup of coffee could activate a CRISPR-based cancer treatment