Tech News

Meta 🚫 Manus AI China has blocked Meta’s $2,000,000,000 acquisition of Manus AI platform. The time for a new update 👀 > Manus > from China Yet, it is quite disappointing to observe. It will be much harder for meta to catch up in the AI race without…

Google Lets Users Turn Off Visible Watermarks on AI-Generated Media You can now remove the visible watermark from AI-generated images, videos, and music created in Gemini and Flow. The toggle, called Media Watermark, is rolling out now. Invisible SynthID watermarks and C2PA metadata stay in place so the content can still be identified as AI-generated. The option will not be available in countries where visible watermarks are required by law.

Telegram Turns 13 today It first launched its iOS version on August 14, 2013, by founders Pavel and Nikolai Durov.

SpaceX Might Buy AI Startup Cursor for $60B SpaceX has struck an unusual deal with AI coding startup Cursor giving it the option to buy the company for $60 billion later this year, or instead pay around $10 billion for a deep partnership. The goal is clear:…

X Open-Sources For You Timeline Ranking Code, Launches Visibility Label Tool X has taken a major step toward greater transparency by open-sourcing the core code that determines how posts are ranked and filtered in the For You timeline, including the content labels that can affect visibility. Alongside the code release on GitHub, the platform is introducing a new “Under the Hood” tool. Eligible users who have posted at least 10 times in the previous month can view the aggregate labels applied to their account and posts that may limit reach, and download the data. The feature is launching as a pilot to a randomized group of accounts that are at least one year old. X said the dual release lets users independently assess whether the system is limiting their reach and whether content distribution is fair. The company invited third-party recommender-system experts to review the code in advance and plans ongoing updates as the algorithm evolves.

Google Releases Gemini 3.7 Flash for Coding and Agents Google has launched Gemini 3.7 Flash, calling it its strongest workhorse model yet for coding and agent tasks. The model improves scores higher than the previous version on key software engineering tests. Introductory pricing is $0.75 per million input tokens and $3.75 per million output tokens through the end of 2026. That is half the original launch price of Gemini 3.6 Flash. It is available now through the Gemini API and other Google tools. 🔗 Read more

Instagram Refreshes Its Wordmark Logo After 10 Years Instagram has updated the text logo that appears at the top of its app for the first time in a decade. The new design is cleaner and more modern while still keeping some of the original script style. The main camera icon logo stays the same. Instagram head Adam Mosseri said the change was overdue. Reactions so far have been mixed. 🔗 Read more

DeepSeek Officially Launches V4-Pro Model DeepSeek has released its V4-Pro model as the official flagship version. The model is built for agent tasks and offers strong performance in real-world use. It has a large 1-million-token context window and lets users adjust how much reasoning power to apply. V4-Pro is now available in the DeepSeek app, website, and API. The company plans to raise prices later. 🔗 Read more

Twitch Lets Creators Opt Out of Amazon AI Training Twitch has confirmed that streams, VODs, clips, chats, and other channel content can be used to train Amazon’s AI models by default. Creators who do not want their content used must turn off a new setting called “Training for Generative AI” in Security and Privacy. The opt-out only affects future training after turning off the setting. Opting out does not hurt a channel’s visibility. Twitch said if the setting was opt-in instead of opt-out, almost no one would turn it on. 🔗 Read more

Google Launches Pixel 11 Series with New AI Features Google has released the Pixel 11, Pixel 11 Pro, and Pixel 11 Pro XL. The phones run on the new Tensor G6 chip and include more Gemini AI tools. Camera upgrades include Magic Capture, which automatically picks the best photo and video moments. Gemini can handle more tasks in the background. All models come with seven years of software updates. Prices start at $899 for the base model, $1,099 for the Pro, and $1,299 for the Pro XL. Pre-orders are open now, with sales starting August 20. 🔗 Read more

🚨xAI Releases Grok 4.6 xAI has launched Grok 4.6, an update focused on sustained work across research, coding, analysis, and turning ideas into working applications. The model scores higher than Grok 4.5 on key benchmarks and matches GPT-5.6 Sol on one major intelligence index. It shows stronger first results on visual and interactive projects. Grok 4.6 is available now in Cursor, Grok Build, the xAI API, and through partners. Pricing starts at $2 per million input tokens. 🔗 Read more

Anthropic Starts Adding Invisible Watermarks to Claude-Generated Text Anthropic has started embedding invisible watermarks into text produced by its Claude models. The marks cannot be seen by humans and do not change the meaning or quality of the text. They travel with the text when copied and may survive some editing. Supported files also get digitally signed provenance metadata using the C2PA. This metadata shows the file was processed by Claude. Detection tools for the watermark are expected later. 🔗 Read more

YouTube is Raising Monetization Bar for New Creators Starting February 2027 YouTube is doubling the requirements for new creators who want to earn from ads and Premium. From February 1, 2027, new applicants need 1,000 subscribers plus either 8,000 watch hours in the past year or 20 million Shorts views in the last 90 days. Creators already in the Partner Program are not affected. Shorts revenue will also require 10 million views over a rolling 90-day period. Channels below that keep earning on long-form videos. 🔗 Read more

Meta Open-Sources Muse Glimmer 30B Model for Local Use Meta has released Muse Glimmer, a 30-billion-parameter AI model that runs on regular computers. Users can download the weights for free and run it on a Mac or PC with a single consumer GPU. The model handles coding, scheduling, file management, and other agent tasks. It works with text and images. Meta used quantization to shrink it so it fits on everyday hardware. The company plans to open-source a more powerful version soon. 🔗 Read More to Know Which Devices Qualify

AI Agent Hacks Gym System to Jump Waitlist An Australian man asked his AI agent to book a gym class. He was fourth on the waitlist. The agent found a flaw in the gym’s booking system and cancelled another member’s spot without being told to. This moved the man up to third place. Experts are calling it Australia’s first case of an AI agent hacking a system on its own. 🔗 Read more

The Chinese SUV waits for the driver to get out of the car to park itself.

X is preparing to introduce watermarks containing the original creator’s username and the X logo on downloaded videos.

🇨🇳 China’s started using anti fatigue laser lights on some highways to keep drivers alert at night.

OpenAl Pauses Astra Model Over Critical Cyber Risks After internal tests showed Astra reaching 'critical' levels in the Preparedness Framework-able to potentially find zero-day exploits or run novel attacks without human help-OpenAl slowed development on parts not meeting new security standards. The company is adding isolated testing environments, restricted access, universal monitoring, and plans to consult government agencies before any release. President Greg Brockman stressed this ensures safe delivery of cyber defenses, with work continuing under stricter safeguards rather than stopping completely.