OpenAI expands o3 capabilities with higher limits and task scheduling

User reports suggest OpenAI has raised the ChatGPT o3 model’s usage quota to 50–100 messages per day. Both o3 and o4-mini now support task scheduling, signaling broader integration ahead of a possible “o3 Pro” tier launch.

🗞 #chatgpt

Google readies native image generation in Gemini ahead of possible I/O reveal

Google is preparing to launch native image generation in Gemini, with backend updates and content rights warnings hinting at an imminent release. A public reveal may coincide with Google I/O, though no official timeline has been confirmed.

🗞 #gemini

OpenAI tests Deep Research Mini tool for free ChatGPT users

OpenAI is introducing “Deep Research Mini,” a scaled-down version of its Deep Research tool, to ChatGPT users, including those on the Free tier, offering topic-specific breakdowns without requiring a premium subnoscription.

🗞 #chatgpt

Google quietly tests Search Agent in Gemini's prompt composer

Google is testing a "Search" agent within Gemini, aiming to expand its role in information retrieval. Integrated under a new "Agents" UI, it auto-verifies answers using Google Search, signaling a shift toward modular, task-specific tools in its ecosystem.

🗞 #gemini

Claude Max plan could bundle Claude Code in push for broader adoption

Anthropic is testing the inclusion of Claude Code in the Max plan and a referral contest offering free subnoscriptions, aligning with its strategy to attract developers and expand user adoption through subnoscription value and viral growth incentives.

🗞 #claude

New NotebookLM feature will let users choose Audio Overview length

Google is testing a NotebookLM feature that lets users adjust audio overview length, supporting brief recaps or extended formats. This aligns with efforts to evolve the platform into a versatile AI research tool for varied professional needs.

🗞 #notebooklm

Meta limits AI features in EU over Llama 4 licensing and data concerns

Meta has launched its AI assistant in the EU, offering only text-based features due to regulatory constraints. Full capabilities remain restricted, with future training on public user content planned, raising privacy concerns among watchdogs.

🗞 #meta

Lovable 2.0 introduces Chat Mode Agent alongside Multiplayer support

Lovable 2.0 introduces real-time collaboration, a security scan for Supabase apps, a smarter Chat Mode Agent, and new pricing tiers, positioning the platform closer to competitors like Replit and Bubble while targeting freelancers and startups.

🗞 #ai

Meta AI will get new Reasoning, Canvas and Connected apps tools at LlamaCon

Meta’s upcoming LlamaCon will showcase major upgrades to its AI web app, introducing new modes like Reasoning, Canvas, Research, Talk, and video creation, as well as expanded voice options, app integrations, and a personalized Memory feature.

🗞 #meta

Alibaba Cloud debuts 235B-parameter Qwen 3 to challenge US model dominance

Alibaba Cloud's Qwen team launched Qwen 3, a family of open-weight language models up to 235B parameters using a mixture-of-experts design. Models are Apache-2.0 licensed and support both cloud and local deployment options.

🗞 #ai

Google expands NotebookLM with Audio Overviews in over 50 languages

Google expanded NotebookLM’s Audio Overviews to over fifty languages, allowing users to generate voice summaries of documents, slides, and trannoscripts. The update improves accessibility, with options for tone, depth control, and offline playback.

🗞 #notebooklm

Anthropic tests custom integrations for Claude using MCPs

Anthropic is testing custom integration support in Claude's web app via the Modular Component Protocol, enabling third-party tools and data connections without local setups, signaling a move toward broader modular and web-based deployments.

🗞 #claude

Google prepares new Gemini AI subnoscription tiers with possible Gemini Ultra plan

Google is preparing new Gemini AI subnoscription tiers, including a possible "Gemini Ultra" plan, with code references suggesting expanded access and usage limits. This move supports its growing push toward a subnoscription-based model.

🗞 #gemini

Meta tests new AI features including Reasoning and Voice Personalization

Meta is developing new features for its AI app, including a "Discover AIs" tab, a "Reasoning" mode for structured explanations, and voice settings like "Focus on my voice" and "Welcome message," aiming to expand customization options.

🗞 #meta

Anthropic launches Integrations and Advanced Research for Max users

Anthropic launches Integrations and Advanced Research mode for Claude, enabling access to remote MCP servers, web, Google Workspace, and business apps. Now in beta for Max, Team, and Enterprise plans, with web search live for all paid users.

🗞 #claude

Microsoft Copilot starts testing Agent Actions and adds native image generation

Microsoft is advancing its Copilot app with native image generation via GPT-4o, code-level hints of an upcoming task automation feature for Windows, and evolving AI personas that support its strategy of integrating utility and branding.

🗞 #microsoftcopilot

Voice mode in Claude app nears launch with web search and file support

Claude’s voice mode is nearing release, offering real-time voice conversations with four voice options, web search support, file uploads, and organized responses. It relies on a push-to-talk system, allowing manual control during usage.

🗞 #claude

Comet AI browser by Perplexity aims to rival Chrome in May

Perplexity AI is set to launch its Comet browser, aiming to integrate AI deeply into browsing workflows for research tasks. With native data access, privacy features, and UI redesign, it positions itself as a challenger to Chrome and traditional search tools.

🗞 #perplexity

Google tests Computer Use tools and Cloud Run hosting in AI Studio

Leaked code suggests Google AI Studio may soon support “Computer Use,” enabling desktop-level control and containerised Gemma 3 deployment via Cloud Run, pointing to Google's broader push into local execution and serverless AI orchestration.

🗞 #aistudio