Perplexity just rolled out Perplexity Assistant on iOS, turning its search engine into an action-taking mobile agent. The app can now draft emails, reschedule meetings, book rides, make reservations, play media, and set reminders right on your iPhone.
Nvidia just made its NeMo microservices generally available, offering end-to-end tools to customize, evaluate, and guardrail enterprise AI agents. It boosts fine-tuning throughput by up to 1.8x and helps firms like AT&T and Cisco tap into continuous data flywheels. Basically, it turns enterprise LLMs into constantly learning digital coworkers.
Microsoft just announced the Microsoft 365 Copilot Wave 2 spring release, introducing an Agent Store alongside two reasoning agents, Researcher and Analyst. Powered by OpenAI deep reasoning models, the agents tackle complex multi-step research and data analysis, while updates also add cross-app search and audio-enabled notebooks. It signals Microsoft's aggressive push into human-agent collaboration.
Descript just announced a new AI-powered agentic video editor, with founder Andrew Mason describing it as the 'Cursor for video.' The team is going all-in on AI agents to streamline editing workflows, and sign-ups for early access are now open.
1Password just launched Agentic AI Security and updated its Extended Access Management platform to manage both human and nonhuman credentials. The tool eliminates hardcoded secrets, enforces least-privilege access for autonomous agents, and includes a dedicated developer SDK. Why it matters: As AI agents take on critical workflows, keeping machine credentials locked down is quickly becoming enterprise security's next big headache.
ElevenLabs just rolled out seamless call transfers between agents in its Conversational AI platform. Different teams inside a company can now build specialized agents in parallel, each equipped with its own tools and knowledge base. Why it matters: Complex voice workflows no longer need a bloated, do-it-all bot.
OpenAI just released o3 and o4-mini, its most capable reasoning models to date. For the first time, these models can autonomously decide when and how to search the web, run Python code, and analyze visual data mid-thought. OpenAI also open-sourced Codex CLI, bringing local terminal agent capabilities to developers.
Docusign just launched AI contract agents integrated into its Intelligent Agreement Management platform to analyze agreements and highlight risks requiring human review. Powered by its proprietary Iris engine, the first agents will debut by the end of 2025 for sales and procurement workflows.
Researchers just launched TextArena on arXiv, a fresh benchmark packing over 57 competitive text games to evaluate how LLMs handle complex agentic behavior like negotiation, deception, and theory of mind. It features real-time TrueSkill ratings, multiplayer support, and model-vs-model battles to stress-test multi-turn planning in dynamic environments.
Microsoft just rolled out an early access preview of computer use in Copilot Studio, letting agents interact directly with desktop apps and browsers like a human. Instead of relying on APIs, agents click, type, and adapt to UI changes in real time. Why it matters: It turns everyday interfaces into tools for agentic automation without custom code.
Anthropic just gave Claude an agentic Research feature alongside direct integration with Google Docs, Gmail, and Google Calendar. Claude can now cross-reference your internal company files with live web searches to deliver cited answers in minutes. It is currently rolling out in beta for paid tiers.
OpenAI just dropped three new models in its API: GPT-4.1, GPT-4.1 mini, and its first-ever GPT-4.1 nano. They feature a 1-million-token context window, jump to 54.6% on SWE-bench Verified coding, and mini slashes costs by 83% while nearly halving latency. It is built squarely for developers building agents and coding workflows, though it won't be available in ChatGPT directly.
OpenAI just published an official prompting guide for GPT-4.1, detailing how to leverage its literal instruction following. By baking in three core reminders around persistence, tool calling, and deliberate planning, OpenAI boosted its agentic SWE-bench Verified score by nearly 20%.
HubSpot just introduced over 200 platform updates at its Spring 2025 Spotlight, headlined by four new Breeze AI agents covering customer support, sales prospecting, knowledge base curation, and content generation. The embedded agents run directly on CRM data, with early support users already automating over 50% of routine tickets.
Google DeepMind just announced it will support the MCP protocol across its Gemini models and SDK. CEO Demis Hassabis called MCP an emerging open standard for the AI agent era, noting Google looks forward to developing it further with the MCP team and the broader industry.
Adobe is rolling out agentic AI across its suite, kicking off with Photoshop's first creative agent dubbed Action Panel. The feature analyzes your image to recommend context-aware edits and triggers over 1,000 actions via natural language. Premiere Pro is also getting agents to piece together rough cuts, while Acrobat adds custom document agents.
Google just introduced Workspace Flows, an agentic AI system that lets users automate multi-step business workflows across apps using natural language and custom Gems. Google is also bringing NotebookLM-style audio overviews and writing coaches to Docs, automated analysis to Sheets, and Veo 2 video generation to Vids. Most new tools are currently rolling out to alpha testers.
Google just introduced Agent2Agent (A2A), an open protocol designed to let AI agents talk and collaborate across different platforms and vendors. Backed by over 50 partners including Atlassian and Salesforce, A2A complements Anthropic's MCP by focusing on multi-agent coordination. A production-ready release is slated for later this year.
Google just rolled out Deep Research powered by Gemini 2.5 Pro Experimental for Gemini Advanced subscribers. The upgrade improves analytical reasoning and synthesis, outperforming rival deep research tools by a 2-to-1 margin in testing, while also supporting podcast-style Audio Overviews of generated reports.
Google just unveiled new remote sensing foundation models alongside Geospatial Reasoning, a research framework that uses Gemini to orchestrate complex spatial analysis. By pairing multi-modal models with satellite imagery and public datasets, agents can map disaster damage and urban infrastructure through natural language queries. Airbus and Planet Labs are among the first testers.
