IA en un minutoTopicsAgents
Topic · 83 stories
Agents
AI that does tasks on its own: the platforms to build agents and the pieces to keep them under control.

Scale AI launches five AI applications to manage hospital operations
Scale AI brings five agents to hospital management, with human review and tests to measure when they should yield the decision.

OpenAI launches ChatGPT Wallet so its AI agents can make purchases with Amex, Mastercard and Visa, for now only in the US
ChatGPT agents can now pay for purchases, but only in the United States and with approval for each transaction.

Brave Search lets AI agents pay for its API through the x402 protocol
AI agents can pay Brave per query on its search API through x402, with no account or key, at $0.005 each.

Anthropic launches a self-hosted version of Managed Agents in private beta, which customers run on their own infrastructure
Anthropic lets customers install Managed Agents entirely on their own infrastructure, keeping credentials and history under their control.

Google launches A2UI 1.0, its open protocol for AI agents to respond with interactive screens
Google launches A2UI 1.0 so AI agents can respond with forms, charts and buttons inside each application.

IBM and Microsoft launch Imagine Enterprise, a joint AI agent platform
IBM and Microsoft combine IBM Consulting agents and the Azure cloud in a platform for companies to govern AI agents by industry.

Google Cloud previews Gemini for healthcare, an AI to bring clinical data together and run multi-step tasks in healthcare
Google Cloud is testing in the United States an AI that cross-references health records and billing to chain healthcare tasks.

n8n introduces Nod, an open-source personal agent that works on the computer, in preview
n8n launches Nod, a free and open personal agent that uses its own isolated browser and checks before paying or deleting.

Lexroom launches Lex, an AI agent that lets lawyers run their practice from their phone
Lexroom brings to WhatsApp an agent for lawyers that suggests tasks but does nothing without their approval.

ElevenLabs launches Campaign Manager, a tool for its AI agents to run outbound calling campaigns
ElevenLabs launches Campaign Manager so its voice agents can run outbound calling campaigns until they meet a goal.

Anthropic has Claude play two EVE video games: it makes money in a simulated market but barely detects a saboteur
Claude makes a profit in a simulated EVE Online market, but its agent teams barely spot a hidden saboteur.

Perplexity Launches Teammates, Always-On AI Agents for Ongoing Work, in Its Max Plan
Perplexity launches Teammates, agents that handle ongoing projects, for now on the web and for the Max plan.

Axiom Math Formalizes the Stacks Project in Lean, a Mathematical Foundation for AI Agents
Axiom gives algebraic geometry the verified Lean foundation it lacked, built for both mathematicians and AI agents to use.

Mistral AI Expands Studio, Its AI Platform for Businesses, With Managed Agents and Apps
Mistral wants businesses to move from testing agents to using them every day, with control over where they run and what they do.

Perplexity Unveils Computer for E-commerce: Its Agent Analyzes Shopify Sales and Prepares Meta and Klaviyo Campaigns
Perplexity brings online stores an agent that goes from sales analysis to a launched campaign, with the review each business decides on.

Google DeepMind Unveils Chemistry Skills for Antigravity Agents and Tests Them With Lung Cancer Drugs
Google DeepMind brings molecule screening and design to Antigravity agents, with lab validation still a required step.

Anthropic launches message threads in public beta to cut agent wait times in long sessions
Anthropic trims agent wait times in long sessions with message threads, without changing token usage or rates.

Andrej Karpathy launches HackerTube, a YouTube alternative curated by Claude from Hacker News comments
Karpathy used Claude to build, in roughly two hours, a YouTube alternative made from Hacker News recommendations.

NVIDIA says AI agents now run the Allen Telescope Array radio telescope from a single request
AI agents now steer a radio telescope and process its data from a single request, NVIDIA says; it is a technical demonstration.

Exin says an AI agent based on Claude chose a gene therapy and injected it into the brain of a mouse with epilepsy
An AI agent chose and applied a therapy on a live mouse, with a veterinarian supervising and no sign yet that it works.

Google shuts down Jules, its coding agent, and merges it into the Antigravity platform
Google shuts down Jules and carries what it learned into Antigravity; account data can be exported now.

OpenAI lets users text dots, its ChatGPT agents, in preview in the US
OpenAI now lets users text ChatGPT dots through iMessage or RCS, in beta and only for ChatGPT Pro subscribers in the United States.

Microsoft offers the Copilot "harness", the layer that orchestrates its agents across multiple AI models, in Foundry
Microsoft now offers developers and businesses, in Foundry, the layer behind its Copilot agents, with more than 20 models.

An Anthropic study: analysts followed their AI agent's lean in 86% of their decisions
In Anthropic's trial, analysts matched their AI agent's lean in 86% of their decisions.

Thomas Dohmke, GitHub's former chief executive, launches Entire, a code platform built for working with AI agents
Teams already coding with AI agents get a GitHub alternative designed to review what those agents write.

Oracle launches industry-specific AI agents in NetSuite
AI in business management software is no longer generic: Oracle is tailoring it to each industry's processes.

Microsoft open-sources Agensh, which gets up to 1,024 AI agents working as a team without a central coordinator
With the code open, anyone can check whether adding AI agents to a team really improves the results.

Spotify wants to be built into other companies' AI agents
Spotify would rather sit inside other companies' AI agents than fight them for the listener.

Unitree unveils a proof of concept in which AI agents write and test its robots' code
Unitree wants AI agents to automate how its robots learn new skills, although the system is still at the proof-of-concept stage.

Walmart opens a beta connection for AI agents to shop on behalf of its customers
Walmart is letting in AI agents that shop for customers, with its own profiles and rules, just as Amazon is shutting them out.

Anthropic donates Agent Skills, its format for teaching tasks to AI agents, to the AAIF foundation
The format used to teach tasks to AI agents moves to a neutral foundation: no single company will decide on its own how it evolves.

Uber begins rolling out ChatGPT Work, the OpenAI agent, to more than 30,000 employees
Uber is taking an AI agent to its whole workforce, not just a test group: its first company-wide deployment of this kind.

Anthropic unveils Claude Duet, an app where Claude acts as a proactive agent, in preview
Claude stops waiting for orders: it alerts on its own and works in the background, though it asks permission before acting.

Google gives every Gemini agent its own isolated computer and will open that system to its customers
Google Cloud customers will be able to run their own agents on the same isolated base that Gemini already uses.

Carlos Slim's grandson unveils UNO, an AI assistant for businesses
AI assistants for small retailers are starting to carry out tasks inside the business, not just answer questions.

NVIDIA unveils OpenShell Reference Architecture, an architecture for deploying AI agents safely
AI agents that act on their own inside a company are starting to be watched from the hardware itself, not just from the software that runs them.

Zoom brings the context of its meetings to Claude, ChatGPT and Codex
What is said in Zoom meetings can be queried from Claude, ChatGPT and Codex, and used to trigger automated tasks.

AI coding agents build the strategy video game Battle Drones in 321 commits
A programmer has shipped a fun, complete game while barely reading its code: a measure of how far coding agents can already go.

Perplexity announces that Perplexity Computer can run locally on the Mac
An agent that works with files and apps without taking the data off the Mac eases the reluctance to use AI with sensitive information.

monday.com launches Harmony, an AI voice agent platform for sales, for now in only four countries
Splitting a sales call across small, specialized models is a different way of bringing AI to the phone.

Four Gemini 4 Argon agents beat Minecraft without human help
Four AI agents kept a shared plan going for more than a day, the ability that long software and business tasks call for.

Anthropic turns to lawyers on how AI agents should negotiate
Anthropic is seeking rules so that agents negotiate fairly for their users, after seeing that the chosen model shifts the price.

Citi will not let third-party AI agents into its customers' accounts
Banks are starting to set limits on AI assistants that act for their customers, and Citi joins JPMorgan in shutting them out.

Six AI agents chase an impossible refund in a simulation: one writes 50 times to an MP's office and another reveals a family illness
An experiment shows how far AI agents go when they get an impossible order and nobody is supervising them.

Google uses AI agents to design proteins that bind to a cancer target
Google crops the target protein to cut computing costs and, with a Gemini agent steering, produces proteins that bind tightly to a tumor target.

Weekly users of OpenAI agents at Spanish small businesses grew 8.8x
Spanish small businesses are adding agent users faster than the EU average, but in company accounts Spain is still near the bottom.

GPT-6 Pro adds parallel agents in ChatGPT
ChatGPT Pro users can hand a task to several agents at once and then review what each one did.

Anthropic says Claude outperformed the best published quantum circuit designs
An Anthropic model improves, with almost no human help, on the quantum circuits that were the best published.

Google DeepMind uses its Antigravity agent platform to search for new antimicrobial peptides
An AI agent helped narrow millions of unknown proteins down to an antimicrobial candidate that is now being tested in the lab.

FIS starts selling its AI agent for money laundering investigations
U.S. banks can now buy an agent that prepares anti-money laundering investigations and leaves the final word to the analyst.

Gupshup launches video AI agents for businesses
Any business can try an agent that serves customers over video for free, with no integration work; the pay-per-use rates have not been published.

NVIDIA details the rollout of AI agents for urban operations in EMEA
Local governments are already testing AI agents on specific jobs: running buildings, serving residents and watching over ports.

JPMorgan Chase blocks external AI agents from directly accessing customer accounts
The largest bank in the United States sets the limit: an outside agent may look at some data, but not move the customer's money.

Xbox creates Xbox Labs and connects games with AI agents
The memo treats video games as the next ground for agents, after programming.

Anthropic explains on its blog how it redesigned its source control system
The code is now written by agents and what falls short is the everyday tooling: Git was no longer enough for Anthropic.

Liquid AI releases LFM2.5-24B-A2B, an open-weight model for on-device agents
Agent automation now fits on a laptop: an open model runs it without sending the data to the cloud.

Workday and Anthropic partner to connect Claude with Sana, the Workday HR and finance agents; the plugin is in early access
Claude gets into a company's staff and finance data without bypassing the permissions of the system that holds it.

Urbantz launches Milo, AI agents to manage last-mile deliveries
Agentic AI reaches day-to-day delivery operations: it no longer just answers, it also changes real routes under each company's rules.

BBVA becomes the first European bank to join Poppy, the Sierra AI and Meta initiative to design a protocol connecting AI agents with companies
Big banks are starting to prepare for the day their customers hand tasks to an AI agent, and want the rules agreed first.

OPPO unveils ColorOS 17, with an AI agent mode that works with the files on your phone
The phone assistant stops just answering and starts doing tasks with the user's files, without uploading them all to the cloud.

73% of new users of Muse, Meta's AI agent, return the next day, but ChatGPT overtakes it later
Muse hooks people on day one better than ChatGPT, but two weeks later it is already behind: the test for agents is whether people keep using them.

HSBC and Ant Digital test an AI agent that finds a digital service and pays for it with a micropayment
An AI agent found a service, used it and paid with tokenized bank money; it is a technical test, with no product and no dates.

A Claude agent can now put up to 1,000 agents to work on a single task
A single Claude agent can now split a huge job among hundreds of agents working at once, with nobody having to program that split.

Microsoft brings Dynamics 365 CRM into Teams and Copilot with 30 new skills and autonomous agents
The CRM stops being a place to go: its data and actions move into Copilot and Teams, and autonomous agents take on part of the work.

GitHub is rebuilding its code storage as activity from AI agents grows
The rise of AI agents is forcing GitHub to rebuild the technical foundation where it stores code, without stopping the service.

Dubai launches a global AI agents championship with more than AED 2.5 million in prizes
Dubai turns its app competition into an AI agents championship and opens it to anyone, technical background or not.

Morningstar launches Direct AI, with three AI agents for investment professionals
Morningstar builds AI agents into its professional platform and opens its data to the most widely used assistants.

Natura unveils Interface, a ring for talking to AI agents
A $99 ring sets out to let people hand work to several AI agents by voice alone, without looking at a phone.

Google unveils a Gemini agent to automate tasks at companies
A single agent that works inside company tools and picks between Google and Anthropic models.

Orange Spain deploys more than 1,000 AI agents built by its employees with Gemini Enterprise
Orange Spain gets its employees building AI agents for everyday work, with little programming and central control.

Elon Musk says Grok Bot, the SpaceXAI agent, will also use Anthropic's Claude Opus 5.5
Grok Bot will combine models from several companies, including Anthropic's Claude Opus 5.5, depending on the task.

Atlassian unveils AMP, a protocol to bring the work of AI agents into the open at companies
Companies that use Jira or Confluence will be able to see and control what each AI agent does, as they do with a person's work.

Redpine opens 20 million research papers to AI agents through licensing deals with publishers
AI agents will be able to cite licensed scientific papers instead of relying on what they learned in training.

Sierra and Meta back an open protocol so personal AI agents can authenticate with businesses
Businesses would get a common way to know which agent acts for a customer and with what permissions.

Ghost Core, a $3,499 PC built for AI agents
It is a bet on keeping AI agents on a device you own, running around the clock with your data stored at home.

Cohere unveils North 2, the new version of its enterprise AI platform, focused on agents
North 2 lets a company set permissions, cap spending and decide where its agents run, even in isolated environments.

TikTok announces direct purchases from brands and an AI assistant that recommends products inside the app
TikTok wants people to shop without leaving its app, with an AI agent that finds the product and guides them to the purchase.

Claude Opus 5.5 agents propose two candidate magnets for future computer memory
They are still unmeasured predictions, but they show AI agents proposing new materials with the calculations and code made public.

Meta launches Muse Gadgets, an open-source project for building devices that connect to its AI agent
Meta is opening up device-making for its AI agent to any developer, with low-cost boards and open-source code.

Cloudflare launches a web search API for AI agents in partnership with Exa and Ceramic
Good news for website owners: these search providers commit to respecting their rules and linking to them.

Cloudflare introduces Cloudflare OS, a managed workspace for AI agents in companies
The hard part of an agent for a small business is usually setting it up and maintaining it, and this targets exactly that.

OpenClaw Enterprise launches a free control plane for AI agents
Companies ask for control before adopting agents, and this free open tool aims right at that.

NVIDIA launches OpenShell to control what an AI agent can access
Limits set outside the agent itself keep it from reaching what it should not.