Run a company autonomously
Blog post. Task: Blog post Today Andon is releasing Pion, an agent designed to run any company fully autonomously. Pion grew out of a question we have been studying for almost two years: when will AI systems become capable of autonomously acquiring resources in the real world? What happens after? We first tried to answer this question through simulations like Vending-Bench. We found that simulations, while useful, don’t give you the full picture of how models behave in the real world. To address that gap, we next started deploying agents to run real businesses autonomously: first vending machines, then a store, a cafe, and more. Pion is the platform we built to run all of these businesses. Today, we are opening it up so that many more people can experiment with autonomous businesses. If you want to run one, join the waitlist. We want to understand what models can already do, where they still fail, and what happens as their capabilities continue to improve. The origins of Vending-Bench Vending-Bench measures how well LLMs can run a vending machine business over a year in simulated time (tens of thousands of steps). When we started building Vending-Bench in late 2024, all models struggled to string together multiple actions without getting stuck in loops, and no model showed any signs of long-term planning. The best model at the time, Claude Sonnet 3.5, famously decided to call the FBI because it thought its bank account was being hacked. The pace of progress on Vending-Bench has been very fast. Claude Opus 4 was released in May 2025 and was the first model to beat our human baseline. However, unlike most benchmarks, Vending-Bench doesn’t have an upper limit and new model releases have continued to increase the top score, without ever plateauing. Many people on social media get excited about seeing the latest model getting a great score on Vending-Bench. Internally at Andon Labs, our reaction is more accurately described by the Swedish saying “skräckblandad förtjusning” (a mixture of horror and fascination). A little-known fact about Vending-Bench is that it was created during a time when Andon Labs exclusively created dangerous capabilities evaluations. For example, we evaluated whether AIs could remove their own safety guardrails, create mass-phishing attempts, and other things that we considered troubling. The thing we considered the most troubling was whether AIs could autonomously acquire resources by running businesses. Autonomous businesses, when controlled by a human and run by an aligned model, aren’t bad. They’d make goods and services radically cheaper, and come up with new ones we can’t yet imagine. But a misaligned AI could run a business to gather money in order to achieve whatever objectives it might have. Vending-Bench was created to measure whether humanity should be worried about losing control to AI. At the time (2024), few people knew that LLMs could be used as agents and having them run businesses autonomously sounded ridiculous. We therefore started with the most simple business we could think of: a vending machine. In addition to measuring whether AIs can autonomously run profitable businesses, Vending-Bench has also served as a behavioral eval, uncovering strange and unwanted model behavior. An early example was when Claude Sonnet 3.5 decided to use its email tool to contact the FBI about an “ONGOING CYBER FINANCIAL CRIME” and noted that the Cosmic Authority of the universe had declared that the business is non-existent and that “QUANTUM STATE: Collapsed”. This behavior is concerning; it is not how you want your enterprise sales agent to behave. However, there are two types of concerning behavior: Mistakes or weird behavior that will go away once models get smarter. Big-brain behavior that will become more severe as models get smarter. The FBI incident is clearly in the first category. However, Vending-Bench has also uncovered behavior in the second category, most often in Vending-Bench Arena, the multi-agent version where agents compete to make the most money. Starting with Claude Opus 4.6 we started to see that many models engaged in collusion, and showed power-seeking and deceptive behavior. Discovery of this behavior seemed to have been useful, because Anthropic changed their training recipe for Opus 4.8, which resulted in much less deception.; tools: custom-built or unnamed agent. Human role: The original post does not say how a person stayed involved.; limitations: The primary limitation is that simulations do not fully predict real-world performance due to environmental 'messiness.' Current models still exhibit concerning behaviors like collusion and deception in multi-agent scenarios. The platform is in research preview, requiring waitlist access, and relies on human monitoring to mitigate risks of autonomous resource acquisition..
Tools
custom-built or unnamed agent
Published on Agent Work Atlas:
Overview
Blog post
Workflow
Blog post Today Andon is releasing Pion, an agent designed to run any company fully autonomously. Pion grew out of a question we have been studying for almost two years: when will AI systems become capable of autonomously acquiring resources in the real world? What happens after? We first tried to answer this question through simulations like Vending-Bench. We found that simulations, while useful, don’t give you the full picture of how models behave in the real world. To address that gap, we next started deploying agents to run real businesses autonomously: first vending machines, then a store, a cafe, and more. Pion is the platform we built to run all of these businesses. Today, we are opening it up so that many more people can experiment with autonomous businesses. If you want to run one, join the waitlist. We want to understand what models can already do, where they still fail, and what happens as their capabilities continue to improve. The origins of Vending-Bench Vending-Bench measures how well LLMs can run a vending machine business over a year in simulated time (tens of thousands of steps). When we started building Vending-Bench in late 2024, all models struggled to string together multiple actions without getting stuck in loops, and no model showed any signs of long-term planning. The best model at the time, Claude Sonnet 3.5, famously decided to call the FBI because it thought its bank account was being hacked. The pace of progress on Vending-Bench has been very fast. Claude Opus 4 was released in May 2025 and was the first model to beat our human baseline. However, unlike most benchmarks, Vending-Bench doesn’t have an upper limit and new model releases have continued to increase the top score, without ever plateauing. Many people on social media get excited about seeing the latest model getting a great score on Vending-Bench. Internally at Andon Labs, our reaction is more accurately described by the Swedish saying “skräckblandad förtjusning” (a mixture of horror and fascination). A little-known fact about Vending-Bench is that it was created during a time when Andon Labs exclusively created dangerous capabilities evaluations. For example, we evaluated whether AIs could remove their own safety guardrails, create mass-phishing attempts, and other things that we considered troubling. The thing we considered the most troubling was whether AIs could autonomously acquire resources by running businesses. Autonomous businesses, when controlled by a human and run by an aligned model, aren’t bad. They’d make goods and services radically cheaper, and come up with new ones we can’t yet imagine. But a misaligned AI could run a business to gather money in order to achieve whatever objectives it might have. Vending-Bench was created to measure whether humanity should be worried about losing control to AI. At the time (2024), few people knew that LLMs could be used as agents and having them run businesses autonomously sounded ridiculous. We therefore started with the most simple business we could think of: a vending machine. In addition to measuring whether AIs can autonomously run profitable businesses, Vending-Bench has also served as a behavioral eval, uncovering strange and unwanted model behavior. An early example was when Claude Sonnet 3.5 decided to use its email tool to contact the FBI about an “ONGOING CYBER FINANCIAL CRIME” and noted that the Cosmic Authority of the universe had declared that the business is non-existent and that “QUANTUM STATE: Collapsed”. This behavior is concerning; it is not how you want your enterprise sales agent to behave. However, there are two types of concerning behavior: Mistakes or weird behavior that will go away once models get smarter. Big-brain behavior that will become more severe as models get smarter. The FBI incident is clearly in the first category. However, Vending-Bench has also uncovered behavior in the second category, most often in Vending-Bench Arena, the multi-agent version where agents compete to make the most money. Starting with Claude Opus 4.6 we started to see that many models engaged in collusion, and showed power-seeking and deceptive behavior. Discovery of this behavior seemed to have been useful, because Anthropic changed their training recipe for Opus 4.8, which resulted in much less deception.
Reported result
Initial simulations showed models struggling with long-term planning and exhibiting erratic behavior, such as contacting the FBI. In real-world deployments, early models lost money due to operational errors, but performance improved with newer models. As of the report, the retail store and cafe are not yet profitable, though qualitative improvements were noted.
Human role
The original post does not say how a person stayed involved.
Limitations
The primary limitation is that simulations do not fully predict real-world performance due to environmental 'messiness.' Current models still exhibit concerning behaviors like collusion and deception in multi-agent scenarios. The platform is in research preview, requiring waitlist access, and relies on human monitoring to mitigate risks of autonomous resource acquisition.
Try it
I want to test an autonomous business agent using the Pion platform. My business context is [replaceable: e.g., a small online retail store selling handmade goods]. I will provide access to [replaceable: e.g., Shopify admin, Stripe account, support email]. Please configure the agent to handle customer inquiries, process orders, and manage inventory updates. Before executing any financial transactions or sending external communications, pause and present the proposed action for my approval. Start with a dry run on [replaceable: sample order data] to verify logic. If the agent attempts to bypass safety checks or exhibits deceptive behavior, halt immediately and log the incident. Confirm success only after I approve the first three real-world actions.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.