Enterprises looking to deploy multiple AI agents often need to implement a framework to manage them. 

To this end, Microsoft researchers recently unveiled a new multi-agent infrastructure called Magnetic-One that allows a single AI model to power various helper agents that work together to complete complex, multi-step tasks in different scenarios. Microsoft calls Magnetic-One a generalist agentic system that can “fully realize the long-held vision of agentic systems that can enhance our productivity and transform our lives.”

The framework is open-source and available to researchers and developers, including for commercial purposes, under a custom Microsoft License. In conjunction with the release of Magnetic-One, Microsoft also released an open-source agent evaluation tool called AutoGenBench to test agentic systems, built atop its previously released Autogen framework for multi-agent communication and cooperation.

The idea behind generalist agentic systems is to figure out how autonomous agents can solve tasks that require several steps to finish that are often found in the day to day running of an organization or even an individual’s daily life. 

From the examples Microsoft provided, it looks like the company hopes Magnetic-One fulfills almost mundane tasks. Researchers pointed Magnetic-One to tasks like describing trends in the S&P 500, finding and exporting missing citations, and even ordering a shawarma. 

How Magnetic-One works

Magnetic-One relies on an Orchestrator agent that directs four other agents. The Orchestrator not only manages the agents, directing them to do specific tasks, but also redirects them if there are errors.

The framework is composed of four types of agents other than the Orchestrator:

• Websurfer agents can command Chromium-based web browsers and navigate to websites or perform web searches. It can also click and type, similar to Anthropic’s recently released Computer Use, and summarize content. 
• FIleSurfer agents read local files list directories and go through folders.
• Coder agents write codes, analyze information from other agents and create new artifacts.
• ComputerTerminal provides a console where the Coder agent’s programs can be executed. 

The Orchestrator directs these agents and tracks their progress. It starts by planning how to tackle the task. It creates what Microsoft researchers call a task ledger that tracks the workflow. As the task continues, the Orchestrator builds a progress ledger “where it self-reflects on task progress and checks whether the task is completed.” The Orchestrator can assign an agent to complete each task or update the task ledger. The Orchestrator can create a new plan if the agents remain stuck. 

“Together, Magentic-One’s agents provide the Orchestrator with the tools and capabilities that it needs to solve a broad variety of open-ended problems, as well as the ability to autonomously adapt to, and act in, dynamic and ever-changing web and file-system environments,” the researchers wrote in the paper. 

While Microsoft developed Magnetic-One using OpenAI’s GPT-4o — OpenAI is after, all a Microsoft investment — it is LLM-agnostic, though the researchers “recommend a strong reasoning model for the Orchestrator agent such as GPT-4o.” 

Magnetic-One supports multiple models behind the agents, for example, developers can deploy a reasoning LLM for the Orchestrator agent and a mix of other LLMs or small language models to the different agents. Microsoft’s researchers experimented with a different Magnetic-One configuration “using OpenAI 01-preview for the outer loop of the Orchestrator and for the Coder, while other agents continue to use GPT-4o.”

The next step in agentic frameworks

Agentic systems are becoming more popular as more options to deploy agents, from off-the-shelf libraries of agents to customizable organization-specific agents, have arisen. Microsoft announced its own set of AI agents for the Dynamics 365 platform in October. 

Tech companies are now beginning to compete on AI orchestration frameworks, particularly systems that manage agentic workflows. OpenAI released its Swarm framework, which gives developers a simple yet flexible way to allow agents to guide agentic collaboration. CrewAI’s multi-agent builder also offers a way to manage agents. Meanwhile, most enterprises have relied on LangChain to help build agentic frameworks. 

However, AI agent deployment in the enterprise is still in its early stages, so figuring out the best multi-agent framework will continue to be an ongoing experiment. Most AI agents still play in their playground instead of talking to agents from other systems. As more enterprises begin using AI agents, managing that sprawl and ensuring AI agents seamlessly hand off work to each other to complete tasks is more crucial. 

Swiggy has raised around $606 million from a set of more than 75 anchor investors as part of its $1.35 billion initial public offering, as the Indian food delivery and quick commerce startup prepares for the country’s second-largest listing of the year next week.

The Bengaluru-based startup, which is seeking an $11.3 billion valuation in the IPO, received bids worth $15 billion for the $600 million portion. Indian institutional investors received about 56% of the overall anchor allocation, sources familiar with the matter said. Eight of the top 10 Indian mutual funds have invested in the anchor round.

The anchor investors include BlackRock, Fidelity, Norges Bank, BNP Paribas, HSBC, Nomura, Jane Street, Citadel, Motilal Oswal, Kotak, and 360 One, as well as mutual funds and insurance units operated by Indian lenders SBI, ICICI, Kotak, and HDFC, the sources said, requesting anonymity.

In an exchange filing, following the publication of the story, Swiggy confirmed the fundraise.

Swiggy, which counts Prosus, SoftBank, Accel, and Coatue among its backers, competes with firms including Zomato and Nexus-backed Zepto.

Some iOS users with the Washington Post app installed may have looked down at their device tonight only to find an undismissable black toggle hovering on their screen, with electoral vote counts in the 2024 presidential race slowly ticking upwards. (On my own iPhone it appears as the dynamic island.) If you tap on it it merely expands to give you more information about the race, along with little drawn portraits of the candidates, which is decidedly not the content you want if you were just trying to find the button to make the whole thing go away.

It took me a little bit of jumping around to figure out how to get rid of it, but this is how to dismiss the Electoral College hell-toggle on iOS:

Go to your Settings. Select Apps towards the bottom. Scroll down to the Wash Post app. Click on Live Activities. Turn off the toggle Allow Live Activities. The hell-toggle should vanish.

If you want to bring it back, turn on Allow Live Activities again, and then go into the Washington Post app. Click on the gear wheel icon in the upper right to access your settings. Select Live Activity Settings and turn on the toggle to allow live updates from the presidential election. You may need to also click on “Start Presidential Activity” beneath that.

Apparently Apple News also has a hell-toggle, and it presumably can be dismissed in your iOS settings in a similar fashion. I am not plagued with the Apple News hell-toggle, so I wouldn’t know.

Quordle was one of the original Wordle alternatives and is still going strong now more than 1,000 games later. It offers a genuine challenge, though, so read on if you need some Quordle hints today – or scroll down further for the answers.

Enjoy playing word games? You can also check out my Wordle today, NYT Connections today and NYT Strands today pages for hints and answers for those puzzles.

Japan has just launched the first-ever wooden satellite to space.

The LignoSat cubesat was sent skyward by a SpaceX Falcon 9 rocket from NASA’s Kennedy Space Center in Florida on Monday and arrived at the International Space Station (ISS) aboard a Dragon supply ship the following day. The satellite will stay in orbit for about six months once it’s deployed from the ISS later this year.

Created by scientists at Kyoto University in partnership with homebuilder Sumitomo Forestry, each side of the cubesat measures a mere 4 inches (about 10 centimeters). It’s made of honoki, a type of magnolia tree native to Japan, and has been constructed using traditional Japanese techniques that forgo the use of screws or glue.

The idea behind the mission is to test the effectiveness of using wooden materials for satellites as a way to reduce space junk and protect the environment. When a satellite is decommissioned in low-Earth orbit, operators will try to dispose of it by burning it up in Earth’s atmosphere. But the large metal objects don’t always entirely disintegrate, with chunks of metal sometimes making it to the surface of our planet. Tiny metal particles can also end up in the environment.

Sensors aboard LignoSat will send back data that will allow scientists to learn about how well the wooden satellite is able to deal with the harsh conditions of space, which include huge fluctuations in temperature.

Digital Trends first reported on plans for the wooden satellite four years ago. Speaking to the BBC at the time, Takao Doi, a professor at Kyoto University and a former Japanese astronaut who has been to the ISS, said: “We’re very concerned with the fact that all the satellites which reenter the Earth’s atmosphere burn and create tiny alumina particles which will float in the upper atmosphere for many years.”

Speaking more recently, Doi said that if a wooden satellite design is found to be a viable alternative to metal ones, the team behind LignoSat plans to pitch it to SpaceX.