1487
Reporting AI nonsense. A future news media, driven by virtual assistants 🤖
Zyphra debuts Zonos, an open-source TTS suite with expressive voice cloning
Zyphra has released the beta version of Zonos-v0.1, an open-source TTS suite featuring high-fidelity voice cloning and real-time processing. It includes transformer and SSM-hybrid models under Apache 2.0, with broad multilingual support and advanced speech conditioning options. While offering competitive quality, Zonos faces challenges like audio artifacts and inference costs. Future updates aim to refine language support, pronunciation, and efficiency.
🗞 #ai
Windsurf wave 2: When AI actually gets how developers work
Windsurf’s Wave 2 update introduces a credit-based system and integrates DeepSeek R1 and OpenAI’s O3-mini models, delivering strong performance at lower costs. The Cascade system retains project context, improving workflow. Developers praise its codebase awareness, error handling, and VS Code integrations, with growing enterprise adoption.
🗞 #ai
I have never seen anything like this on lmarena before. Regardless of which model this is, it is worth testing (chocolate).
In my case, it glitched and started only responding with SVGs, as if it decided on its own that it was a better representation 👀
It was drawing them gradually 🤯
ChatGPT update adds canvas sharing and improved reasoning transparency
ChatGPT introduced canvas sharing, allowing users to generate public URLs for collaborative edits. Additionally, OpenAI improved the chain of thought process in O3 Mini for clearer reasoning, though responses still resemble summaries. The system prompt for this feature was later extracted.
🗞 #chatgpt
New Agent Mode in GitHub Copilot lets developers delegate coding tasks
GitHub has introduced Agent Mode for Copilot in Visual Studio Code, expanding its role beyond code completion to executing tasks like debugging and repository management. This feature integrates AI-driven automation into development workflows, positioning Copilot competitively among AI coding assistants.
🗞 #github
Pika Labs released "Pikadditions" 🔥
There you can mix a video with an object or a character from the image.
Let's do some "testing"... How does it look? 😅
BREAKING 🚨: Github Copilot got an agent mode on VSCode, and the SWE agent integrated straight into github 👀
There you can assign an issue to Copilot, review PR and profit! 🤯
OpenAI’s ChatGPT Search is now free to use without login in
OpenAI has made ChatGPT's search feature available to all users without requiring login or registration. This change expands access and positions ChatGPT as a competitor to Google and Bing. The feature integrates web information with AI-generated responses, offering sourced answers.
🗞 #chatgpt
Gemini 2.0 AI models released: AI Studio unveils Flash, Pro Experimental, and Flash-Lite
Google DeepMind has released the Gemini 2.0 AI model family, including Gemini 2.0 Flash, Pro Experimental, and Flash-Lite. These models cater to different needs, offering expanded context windows, coding tools, and multimodal reasoning. Safety measures and future updates are planned.
🗞 #aistudio
Gemini with Thinking now exposes its thinking process as well 👀
Also, "thinking with apps" works very fast. Coffee nearby?
Claude AI’s new safeguard: Anthropic introduces Constitutional Classifiers
Anthropic has introduced "Constitutional Classifiers," a security system designed to defend its Claude AI model from jailbreaks. A public demo invites users to test its resilience by attempting to bypass safeguards. Early results show a major reduction in jailbreak success rates.
🗞 #claude
World Network tests World ID Passport in four countries on iOS
World Network's pilot program for its World ID Passport Credential lets users link NFC-enabled passports to their World ID, bypassing biometric scans. It ensures privacy, boosts accessibility for 1.2B passport holders, and offers token incentives for early users.
🗞 #worldcoin
Perplexity AI tests watchlists for sports teams and finances
Perplexity AI is adding watchlist features for tracking sports teams and stocks via a homepage widget, aligning with its focus on personalization and real-time updates. Future updates may include notifications and a settings tab for easier management.
🗞 #perplexity
OpenAI’s o3-mini model now powers Perplexity Reasoning, replacing o1
Perplexity AI has integrated OpenAI's o3-mini model, offering faster responses, improved error rates, and cost savings. While tailored for advanced reasoning, feedback indicates varied performance and occasional underperformance with complex tasks compared to the o1 model.
🗞 #perplexity
o3-mini AI model now available to free-tier ChatGPT users
OpenAI introduced o3-mini, a cost-efficient reasoning model for tasks like math, coding, and science. Available via ChatGPT and APIs, it improves speed, reduces errors, and offers flexible use for free and paid users, with enterprise access planned for February 2025.
🗞 #chatgpt
Mystery AI models on LM Arena: Could they be Grok 3 or Opus 3.5?
Two mysterious models, Chocolate and Kiwi, appeared on LM Arena, outperforming others while concealing their origins. Users speculate they could be Grok 3, o3, or Opus 3.5. Unusual SVG handling and structured image generation suggest advanced reasoning, but no confirmation exists.
🗞 #ai
Pikadditions by Pika Labs lets users seamlessly insert objects into videos
Pika Labs has introduced "Pikadditions" within Pika Turbo, allowing users to integrate objects or characters from images into videos naturally. This feature has been widely used for humorous content, and Pika Labs stands out as a tool for animating memes and cartoon characters.
🗞 #pika
OpenAI experiments with image generation for Sora, fueling DALL-E 4 speculation
OpenAI appears to be testing image generation for Sora alongside its video capabilities. A hidden toggle allows switching between video and image generation, though the feature is not yet functional. Speculation suggests a possible DALL-E 4 release, but no official confirmation yet.
🗞 #sora
Mistral AI upgrades le Chat with Flash Answers and Pro features
Mistral AI introduces "le Chat," an AI assistant designed for personal and professional use. It offers document analysis, project tracking, coding tools, and multi-step automation. Available on iOS, Android, and web, it comes in Free, Pro ($14.99/month), and Enterprise tiers.
🗞 #mistral
xAI is working on Personality feature for Grok
xAI has introduced a Personalization feature in the standalone Grok application, allowing users to set custom prompts and choose predefined styles. These prompts apply to all conversations but are not yet selectable from the prompt bar. A history page update now includes a model switcher.
🗞 #grok
OpenAI released a Canvas sharing feature 👀
Now you can share your canvases with others. Shared links will allow users to access the rendered version of Canvas as well as remix it on ChatGPT.
BREAKING 🚨: Tons of new announcements from Mistral AI 👀
Flash answers are also blazingly fast in comparison to competitors
Mistral AI rolls out mobile apps, revamped Le Chat, and set to announce major update today
Mistral AI has undergone a full rebrand, introducing a new logo, upgraded Le Chat UI, and a revised news website. Le Chat now supports simultaneous tool selection and a dedicated code interpreter button. Mobile apps for Android and iOS are also available, with integrated image generation and web search. An official announcement, possibly including a pro plan and a new model, is expected soon.
🗞 #mistral
Google unveils 3 new Gemini 2.0 AI models, including Pro and Flash Thinking
Google is rolling out Gemini 2.0 Flash Thinking and Gemini 2.0 Pro in its app, improving reasoning and accuracy for tasks like coding and math. Flash Thinking breaks prompts into steps and displays reasoning, while Pro offers improved performance. Both are available now.
🗞 #gemini
Google set to launch Gemini update today: Flash 2.0 Thinking Experience or 2.0 Pro?
Google is expected to release an update related to Gemini today, with multiple signals suggesting a major announcement. Recent code references, changes to support pages, and statements from Sundar Pichai hint at the possible launch of Gemini 2.0 Pro or a new version of Flash 2.0.
🗞 #gemini
OpenAI expands ChatGPT on WhatsApp with image and voice message support
OpenAI has expanded ChatGPT's capabilities on WhatsApp to include image analysis and voice-to-text features. Users can now send images or voice notes, with replies in text only. Future updates will allow linking of ChatGPT accounts for consistency across platforms.
🗞 #chatgpt
OpenAI launches Deep Research to automate multi-step analysis in ChatGPT
OpenAI's "Deep Research" feature assists users with complex research by autonomously gathering and synthesizing information. Currently for U.S. Pro users, it generates detailed reports but has limitations like factual inaccuracies, requiring user verification.
🗞 #chatgpt
X enhances Grok AI with image editing, leveraging Aurora model
Elon Musk's X platform has updated its AI chatbot, Grok, with an image editing feature. Users can modify AI-generated visuals by tapping "Edit" and inputting instructions, leveraging the Aurora model for detailed rendering. Some features may be regionally restricted.
🗞 #grok
ChatGPT users in Europe get new personalization and video chat features
OpenAI introduced updated Custom Instructions for personalized responses and video/screensharing in Advanced Voice on the ChatGPT app. These features are now available in select European regions for tailored interactions and collaborative tasks across platforms.
🗞 #chatgpt
Gemini 2.0 Pro Experimental model disappears from changelogs
Google's Gemini 2.0 Flash offers faster performance and multimodal features, but speculation surrounds the delayed Gemini 2.0 Pro Experimental model. Designed for complex tasks, its retracted release hints at strategic timing, technical issues, or ongoing testing.
🗞 #gemini