1487
Reporting AI nonsense. A future news media, driven by virtual assistants 🤖
Pika Labs unveils Pika 2.1 AI model with 1080p video quality
Pika Labs has launched Pika 2.1, an AI model offering 1080p video resolution, improved realism, extended clip lengths, advanced motion physics, and enhanced text-to-video capabilities. It supports seamless customization for animators, marketers, and content creators.
🗞 #pika
Perplexity launches US-hosted DeepSeek R1, hints at EU hosting soon
Perplexity introduced DeepSeek R1, a powerful reasoning model with transparent reasoning traces for deep web research. The release includes a "Pro Search" mode selector, desktop notifications, increased usage limits, and plans for future EU hosting options.
🗞 #perplexity
Exclusive: Grok-3 model from xAI spotted ahead of its anticipated release
Grok-3, xAI's upcoming AI model, has been tested on limited platforms, outperforming predecessors and rivaling models like O1 and DeepSeek R1. With public access expected next week, comparisons with O3-mini and Gemini 2.0 Pro remain highly anticipated.
🗞 #grok
Moonshot AI launches Kimi k1.5 with free real-time search and file analysis
Moonshot AI introduces Kimi k1.5 on the web, offering free tools like web searches, multi-file analysis, advanced reasoning, and improved image understanding. With English support in progress, the platform aims to compete with major AI solutions globally.
🗞 #ai
Google working on Code Artifacts for Gemini, inspired by Canvas and Artifacts
Google is advancing Gemini's features, including Code Artifacts for in-sheet code edits, Scheduled Prompts with a 5-task limit, GEM tagging for simplified access, and an Enterprise Mode for privacy management. Gemini Pro 2 and related updates are anticipated next week.
🗞 #gemini
Mistral AI introduces interactive table editing in Le Chat
Mistral AI's Le Chat introduces interactive Markdown tables rendered on a canvas, allowing users to edit, sort, and manipulate data. Additional updates include web search, advanced document/image processing, and a canvas for ideation, boosting productivity tools.
🗞 #mistral
Pika Labs is about to challenge Google Veo 2 with upcoming version 2.1 video model
Pika Labs announced its upcoming video generation model, version 2.1, currently exclusive to trusted testers. Early access indicates high performance in text-to-video and image-to-video features, poised to rival industry leaders like Google's Veo 2 in quality.
🗞 #pika
Google keeps polishing "Scheduled Prompts" feature and working on tagging for Gems 💎
* These features are still in development
** It may take another forever until the release
Copilot Think Deeper opens to free users, hinting at possible o3-mini release
Microsoft's Think Deeper feature for Copilot is now available to free users, with a limit of three messages per day. Previously exclusive to Pro users, the feature's model remains unclear, sparking speculation about upcoming o3-mini releases and potential announcements.
🗞 #microsoftcopilot
BREAKING 🚨: Microsoft is working on Characters for its Copilot 👀
Characters will be able to speak and react to mouse clicks. So far, there are two characters: "Hikari" and "Mica".
* This feature is still in development
Mistral rolled out a new feature called "schteumeul" which allows rendering markdown tables with Canvas.
There you can edit data, add or delete rows and apply different filters 👀
Gemini AI expands with multi-extensions and file uploads for Gemini Live
Google's Gemini AI now supports images, files, and YouTube content in Live conversations on Galaxy S24/S25 and Pixel 9 devices. Updates include expanded app actions, improved accessibility, and advanced reasoning via the Gemini 2.0 Flash model with a vast context window.
🗞 #gemini
DeepSeek R1 now combines real-time web data with advanced reasoning
DeepSeek R1, a reasoning-centered AI model, now merges real-time web browsing with its reasoning abilities. Users can access this feature via the DeepSeek API or platform, while the company focuses on open-source growth and accessibility through smaller model versions.
🗞 #deepseek
Perplexity AI launches Sonar API as a real-time generative search tool
Perplexity AI's Sonar API is a real-time AI search tool available in Base and Pro tiers. It offers cost-efficient, customizable, citation-backed search solutions, with early adoption by Zoom and others, targeting industries like healthcare and finance with compliance needs.
🗞 #perplexity
OpenAI’s new ChatGPT feature blends memory with search functionality
OpenAI is testing a memory-integrated search feature in ChatGPT, allowing it to use stored user data for context-aware responses. Though not yet public, this raises personalization capabilities and sparks discussions on privacy and targeted advertising implications.
🗞 #chatgpt
Meta AI introduces memory feature on its apps in US and Canada
Meta has updated Meta AI with new personalization and memory features, allowing tailored suggestions using Facebook and Instagram data. Built on Llama 3.3, it integrates Bing search, multimodal queries, and privacy controls, available across Meta platforms.
🗞 #meta
Alibaba rolled out Qwen Chat v0.2 and Qwen2.5-1M model
Alibaba unveiled Qwen Chat v0.2, integrating real-time web search, text-to-video generation, and text-based image creation. Alongside, the Qwen2.5-1M models introduce 1-million-token context handling with faster processing, advancing large-scale AI capabilities.
🗞 #ai
New social tab on World App beta enables username search and payments
The World App beta for Android introduced features like a Social tab for username search and money transfers, with a chat feature in development. Updates include Web3 app management, but account transfer issues persist. Rumors link Sam Altman to digital credential efforts.
🗞 #worldcoin
OpenAI rolled out Canvas support for o1 model with React and HTML rendering
OpenAI expanded Canvas to the o1 model, added HTML/React rendering for all users, and introduced macOS app support. o3 mini will offer hundreds of daily queries, and new agent tools are expected, potentially linked to "software making software" initiatives.
🗞 #chatgpt
xAI develops voice mode for iOS app with Grok and ElevenLabs models
xAI is developing a hidden voice mode for its Grok iOS app, featuring voice model options from Grok and 11 Labs. It includes a model selector, multiple voice choices, and real-time data access, aiming to rival ChatGPT and Perplexity in the voice assistant market.
🗞 #grok
xAI is working on map-enabled location cards for Grok AI on X
Grok AI, developed by xAI, is adding location cards with map views on X, aligning with Musk's vision of a multifunctional app. This feature supports real-time, location-focused queries and complements Grok's existing text, image generation, and multilingual capabilities.
🗞 #grok
Copilot’s next evolution? AI Characters and Shopping features spotted in testing
Microsoft is developing character-driven features for its AI Copilot, introducing "Hikari" and "Mika" characters with themed backgrounds. These additions focus on visual personalization while retaining functionality. A shopping assistant feature is also in progress.
🗞 #microsoftcopilot
xAI prepares to roll out custom instructions for Grok on X platform
Grok, xAI's assistant for the X platform, is set to introduce custom prompts, allowing users to tailor responses in tone, depth, and focus. This update aims to broaden its application for personal and professional needs, showcasing xAI's focus on personalization.
🗞 #grok
Perplexity AI launches new assistant feature exclusively for Android users
Perplexity AI has introduced Perplexity Assistant for Android users, integrating reasoning, web search, and apps for task management. Supporting 15 languages, it enables actions like booking rides, setting reminders, and contextual multitasking via the Perplexity app.
🗞 #perplexity
OpenAI launches Operator, an AI agent for autonomous web tasks
OpenAI's "Operator" is an AI agent capable of autonomous tasks in web browsers, mimicking human actions through a Computer-Using Agent (CUA) model. Currently a research preview for ChatGPT Pro users, its rollout prioritizes safety, privacy, and iterative improvements.
🗞 #chatgpt
Gemini 2.0 Flash Thinking Exp-01-21 debuts with 1M context and code execution
Google's Gemini 2.0 Flash Thinking AI model update, Exp-01-21, introduces a 1M token context window, native code execution, longer outputs, and fewer contradictions. With improved benchmarks in math, science, and multimodal reasoning, it's now available free via AI Studio and API.
🗞 #aistudio
xAI developing new model selector for Grok, hinting at Grok 3 launch
xAI is developing a model selector for its Grok platform, currently limited to "grok-2-latest," with plans to expand options, including Grok 3. This move supports user customization and positions Grok as a competitor to leading AI models while broadening its reach.
🗞 #grok
Kimi k1.5 by MoonshotAI achieves SOTA benchmarks in reasoning
MoonshotAI introduced Kimi k1.5, a multi-modal LLM excelling in reasoning across text and vision. It supports 128k token contexts, outperforms competitors in benchmarks, and integrates advanced RL techniques for robust problem-solving in mathematics, coding, and visual tasks.
🗞 #ai
Open-Source DeepSeek-R1 challenges OpenAI’s o1 in advanced benchmarks
DeepSeek has launched DeepSeek-R1, an open-source AI model rivaling OpenAI's o1 in performance. With 671B parameters, advanced reasoning capabilities, and an MIT license, it supports broad use cases via web, API, and six scalable variants for developers and researchers.
🗞 #deepseek