AI News: OpenAI Finally Released What We Asked For

Thinking Machines Labs and Their New AI Model

• A new AI model from Thinking Machines Labs has been released, showcasing impressive demos, including real-time translation capabilities that can speak over someone in a different language without waiting for them to finish, and this new model is a significant development 10s.
• Thinking Machines Labs was founded by Mera Marott, the former CTO of OpenAI, who left the company after Sam Altman's brief departure and subsequent return, and Mera Marott brought several people from OpenAI, as well as some from DeepMind and possibly Anthropic, to join her new company 2m6s.
• The demos shared by Thinking Machines Labs include features such as web search, artifacts, and the ability to interrupt or wait for someone to finish speaking, depending on the context, and these features are designed to make human-AI interactions easier 4m42s.
• The model can also recognize when someone is done speaking and provide a summary or count of specific items mentioned, such as animal names, and it can even give instructions or warnings when necessary, like advising against taking elderly parents mountain biking 6m15s.
• Additionally, the model can watch and provide feedback on someone's behavior, such as alerting them when they start to slouch, and it can interrupt conversations when it thinks it's necessary to redirect the topic or provide a warning, demonstrating its ability to engage in complex interactions 8m30s.

Features and Capabilities of the Thinking Machines AI Model

• The model's ability to reframe language in real-time is also showcased, where it can take someone's negative statement and rephrase it into uplifting and professional language instantly, highlighting its potential applications in various fields 10m50s.
• OpenAI has released a new AI model that is aware of time and can keep track of conversations, allowing users to set time limits and receive reminders, and it can also perform simultaneous tool calls, such as searching the web or generating UI, while conversing with the user 10s.
• The new AI model is not publicly available yet, but demos are available, and a limited research preview is expected to be released in the coming months, with a wider release later this year 2m6s.

OpenAI's New AI Model and Features

• Crusoe has a product called managed inference, which is a high-performance inference platform designed to run large-scale AI workloads with low latency and high throughput, making it ideal for scaling AI apps 4m30s.
• Crusoe's managed inference platform includes a technology called memory alloy, which retains and reuses context across requests, reducing lag and compute burn, and making it suitable for AI apps with long prompts or context-heavy systems 6m10s.

Crusoe's Managed Inference Platform

• OpenAI has also released a new feature that allows users to access Codeex from their phones, using codecs to manage and control their computer-based Codeex sessions remotely 8m40s.
• The Codeex mobile feature allows users to set up their phone to control their computer, scan a QR code, and connect to their computer, enabling them to access and manage their Codeex sessions, including starting new chats and inspecting their vault 10m50s.

OpenAI's Codeex Mobile Feature

• The ability to remotely access a computer from a phone is a game changer, allowing users to work on code, check progress, and respond to questions from their phone, making it convenient for tasks such as coding with codecs 10s.
• OpenAI has released Daybreak, which seems to be their answer to Anthropic's Mythos, but with a different approach, where users can request OpenAI to scan for security vulnerabilities, rather than having direct access to the model 2m6s.
• Anthropic has rolled out a new feature called Agent View in Claude Code, which consolidates multiple agents into one screen, providing a cleaner layout for users who work with the command line interface and spin up multiple agents at once 4m6s.

Crea AI's New Image Model and Features

• Crea AI has released a new image model called Crea 2, which offers functionality and controllability similar to Midjourney, allowing users to control the level of stylization, use multiple images, and create mood boards to generate images that fit a specific style 5m30s.
• The Crea 2 model also allows users to load their own images, generate more that match their style, and create mood boards to analyze and generate images that fit a specific mood or style, with access currently available to Max or business plan users 7m30s.
• OpenAI has released a feature that allows users to create a mood board with pre-made or custom images, and it can generate images based on the mood board's aesthetic, such as a purple-themed board with high-contrast and saturated purple palettes 10s.
• The feature can be used to generate specific images, like a baseball player hitting a home run with a purple aesthetic, and it can produce multiple images that match the desired style 42s.

Google's Android and Gemini Button Updates

• Google has announced updates to Android, including a new feature that allows users to take a picture of a flyer and automatically generate a tour booking in Expedia, or reserve parking with Spot Hero, using the Gemini button on the Chrome browser 2m6s.
• The Gemini button on Android will have the context of the web page and allow users to fill out forms with a single tap, using stored information like passport details and driver's license details 2m6s.
• Google is also upgrading the spoken text feature on Android, which will clean up spoken text by removing filler words like "h" and "ums", and allow users to correct mistakes easily 2m6s.

Google's New Google Book Laptop and AI Features

• Additionally, Google has introduced the Google Book, a new laptop that comes with a new operating system designed for AI, which builds upon the Chromebook and includes AI features like those found on Android phones 2m6s.

OpenAI's New Mouse and Voice Interaction System

• OpenAI has introduced a new way of interacting with devices, reimaging the mouse pointer to allow users to perform tasks without typing, by highlighting and clicking on items, such as adding ingredients to a shopping list, editing images, and merging cells in a document, all with just mouse clicks and voice commands 10s.
• The new system allows users to highlight text or objects and give voice commands to move or edit them, without needing to type any prompts, making it a combination of mouse movements and voice commands to cause actions to happen 42s.
• The system also uses head tracking and eye tracking, allowing users to generate images based on a menu and style from another image, and to give commands to move objects around, making it feel like a more advanced and futuristic way of interacting with devices 2m6s.
• The new feature is expected to be part of the new Chromebook laptop experience and may be available on other devices in the future, potentially allowing users to interact with devices by just dragging their fingers around and giving voice commands 2m6s.

Anthropic's Subscription Model Changes and User Reactions

• Enthropic has announced an increase in the weekly limits for Claude Code by 50% until July 13th, but also announced changes to the Claude subscription model, which will give users a certain amount of credits per month and then charge them at the API rate once the credits are used up, a change that has been met with criticism from some users 4m30s.
• The new subscription model has been described as a "massive nerf" by some users, who claim that the new credit system is expensive and will result in less usage than before, despite the company's efforts to present it as a positive change 6m0s.
• Users are concerned that the new plans from Anthropic will burn out in a few hours of heavy use, making them useless for serious development work, and will result in unexpected billing at the normal API rate 10s.
• Despite this, Anthropic has reportedly beaten OpenAI in business adoption, with a 3.8% rise in adoption to 34.4% of businesses, while OpenAI's adoption fell 2.9% to 32.3% 1m30s.

Anthropic's Industry-Specific AI Solutions

• Anthropic has announced Claude for the legal industry, releasing a range of connectors, plugins, and other tools specifically designed for the legal sector, following similar releases for the financial services, healthcare, design, and cybersecurity industries 2m6s.
• Claude for small businesses allows users to toggle on the service and access pre-built agents that work across various industries, including finance, operations, and customer service, and can automatically connect tools like PayPal and QuickBooks 2m50s.
• A user of Anthropic's Claude was able to recover access to their Bitcoin wallet, which had been locked for over 11 years, by dumping their old computer files into the system and using it to decrypt the wallet 4m10s.

Meta's AI Updates and New Features

• Meta has added an incognito chat feature to WhatsApp, allowing users to have private conversations with their Meta AI that are not saved and are processed in a secure environment 5m20s.
• Meta's large language model, Muse Spark, is being rolled out with faster voice responses, smarter AI glasses, and new shopping and conversation features, making Meta AI more capable and useful in everyday life 6m0s.
• Notion has released an update for developers, including a Notion CLI that allows users to interact with Notion directly from their terminal 7m10s.

Notion's Developer Tools and API Integrations

• Notion has introduced various features, including workers, database sync, agent tools, web hook triggers, external agents API, and Notion agents SDK, allowing users to build apps or work with Notion directly from their terminal, and also enabling external agents like OpenClaw or Hermes to interact with Notion on their behalf 10s.
• The original founders of Dig, Kevin Rose, and one of the original founders of Reddit, Alexis Ohanian, have relaunched Dig, a social upvoting and downvoting site, which utilizes AI to analyze trending topics on X based on the top 2,000 voices in the world of AI, providing a platform to stay up-to-date with AI news 1m30s.

Relaunched Dig AI News Platform

• A new platform, Dig AI, available at di.gg, surfaces trending AI news, including the most popular stories and GitHub repositories gaining traction, making it easier to stay current with AI developments 2m30s.
• An Instagram account, Chat GPT Tricks, posted an image on X, claiming it was generated using AI in the style of a Monae painting, but it was actually a real Monae painting, and the post received negative comments from users explaining why the art was inferior, highlighting the misconception about AI-generated art 4m10s.

AI Art Misconceptions and Open-Source 3D Generation Tools

• World Labs has developed a fully open-source tool that generates 3D environments, meshes, physics, lighting, and audio from a single input image, allowing users to interact with the generated objects and environment, and this tool is available on GitHub and can be used inside Claude Code 6m20s.
• A new GitHub repository has been released, allowing users to install and prompt AI models via cloud code, which is an exciting development for AI enthusiasts 10s.
• Rivian has introduced a new Rivian assistant and Riven unified intelligence, an AI system integrated into Rivian vehicles that can perform various tasks, such as controlling heated seats, reading text messages, and providing vehicle diagnostics 1m42s.

Rivian's AI Assistant and Vehicle Integration

• The Rivian assistant can also serve as an encyclopedia for the vehicle, allowing users to ask questions and receive step-by-step instructions, and it is likely that other car manufacturers will adopt similar AI features in the future 2m6s.
• Figure Robotics has been live-streaming a robot sorting packages for over 34 hours, with the robot having sorted 43,000 packages, demonstrating the potential of automation in logistics 4m10s.

Figure Robotics and Automation in Logistics

• Google IO is scheduled to take place next week, and there are rumors about potential announcements, including a new version of Gemini, a more affordable and faster AI model capable of performing 92% of what GPT 5.5 can do 6m30s.

Google IO Rumors and Expected Announcements

• A new Gemini Spark agent is expected to be revealed during Google IO, which will work as a 24/7 assistant that can learn from user behavior and interact with connected apps and skills 8m0s.
• There is also speculation about the potential announcement of a new version of Google glasses, which could feature displays and various other features 9m40s.
• Google is expected to provide an update on previously demoed technologies, although a launch is not anticipated, and this update is expected to happen at Google IO, which is taking place next week 10s.
• There will be various updates on Android and Google Book, as Google typically saves major announcements for the IO event, and these updates will likely be discussed in an end-of-week news video 1m5s.

Updates to the Chat GPT App and Weekly AI News Roundup

• Thinking Machines has released impressive updates, and the chat GPT app now features codecs that allow for coding on-the-go, eliminating the need to be at a computer, which is a significant development 2m6s.
• The goal is to provide a weekly roundup of the most important AI news, filtering out unnecessary information and hype, to make it easier for viewers to stay informed without having to constantly follow the latest developments 3m30s.
• Viewers can expect an end-of-week video that summarizes the key takeaways from Google IO, as well as any other relevant AI news from the week, and subscribing to the channel will ensure they receive these updates 4m20s.