Interesting to see these experiments. This is early but imagine in few years there will be companies mostly run by agents with a light overview from a human operator. What then happens to scaling of the bussinesses? I would assume, just like today anyone can vibe code an app, there will be vibecoded bussinesses. Maybe its time to start building infrastructure for these bussinesses instead, its a non existent market yet, but give it a few years.

Meanwhile, I can't even get Astra to consistently re-use the same font-size across all of my HTML page headings+subheadings.

There's just no world where this actually results in a stable, respected business. It will be death by a thousand bad impressions, mistakes and oversights. Yeah, you can automate everything. That doesn't mean it's being done well.

That's not to say this might not be something feasible in 2-3 years from now, but we're still a long ways away.

I am in the process of attempting to have AI run my business. I'm actually making very good progress, but it's happening in pieces - I document some task and have it take over, or I give it something to handle while staying in the loop and providing feedback until it's in good shape. At this point it's handling large swathes of my operations, marketing and finance.

The experience of getting it there makes me pretty skeptical of the idea of a general business agent like this. First, because I still find myself having to review some categories of work for errors. These are decreasing over time, but they're still there. I fully believe that as models get better, errors will decline, but I am somewhat surprised to see some of the errors current models make given their intelligence. A common set is having Claude take some product photos and turn them into lifestyle ones using ChatGPT via Claude in Chrome. It's a pretty well-honed workflow at this point, but it'll still return images where the product is obviously not correct and seemingly not notice them.

Anyway, even if the agents are "perfect" in terms of their ability to execute tasks, there's still just an enormous amount of nuance and context in each business that takes a ton of time to convey. I've been at this for a couple of years now, and I'm still clarifying things. Now maybe an agent starting a business from scratch would have a better time since it's not inheriting all of this, but to have one run an existing business requires a very extended handoff, even if the agent is objectively amazing at all aspects of running a business.

There was surprisingly little information on how they actually do this, but we run our business with a large number of, what we call "AI employees" in addition to regular employees, and they act in interesting ways. We've been building out orchestration tools to handle this, and we'll probably do a write-up or blog post on it soon. May even open source some of it.

The preview is that the problem with most agents (and this includes frameworks like Grokbot and Openclaw and Hermes) is that for many of them, they're black boxes. They say they learn or improve, but it's a black box in what they do. Getting agents to reliably do things is hard, and getting agents to build out software tools to help themselves improve and do better over time is also hard.

Our approach at a high level is pretty simple: every single AI employee is a standalone GitHub repo that shares some characteristics, but we direct them to build as much software as possible to make their goal as easy and reliable to manage as possible. Then we have a shared communication layer for bots across the company to interact with humans and AI. We have decided to organize these like departments similar to the way you might hire out humans. I'm not 100% sure if that's the best approach, but I will say it's been easier for people to understand because they're more mentally easily able to traverse the bot org chart if it somewhat reflects a traditional business org chart.

Each of these AI employees has specific sets of goals and KPIs, instructions that they manage the business with manager bots. We have layers of management, which we actually have found helpful. We also run different bots with different models and harnesses, and some using different models and harnesses to check the work before anything can get done, along with lots and lots of testing.

Every single time, actions have a massive amount of tests based off of previous failures to prevent failures in the future. Sorry for rambling. I do think this is a very interesting space. I didn't see anything interesting in Pion that was public on this website, but I do anticipate that more companies will be "AI and software first," as in the substrate of the company is basically a software application powered by autonomous agents, with humans as a fallback.

Correct me if I'm wrong but a human will be able to cut through the noise with quirky advertising, new/novel distribution methods, etc.

Most of the bottleneck in business isn't building the things or sourcing, it's mostly advertising/sales.

These specifically require doing something unique or interesting.

Sure, maybe LLMs can help with fulfillment or operations, but distribution still remains the hard part.

In cstross's Accelerando (2005) [1], there is an autonomous group of shell corporations all-but-DDOSing Delaware by spawning new companies and dynamically changing corporate structure.

Back then, I believe, Stross had already conceptualized "corporations as slow AI", (which is a separate concept from above).

[1] which you can read for free online https://www.antipope.org/charlie/blog-static/fiction/acceler...

I hope this goes better than the prior blog posts from AndonLabs have gone! I was just reading there 2023 article about a coffee shop experiment that was easily persuaded to give away free pastries.

Now that I think about it anyone want to open up a croissant selling shop near me using Pion? I promise to only ask for free chocolate croissant occasionally.

how does this work from a liability standpoint? seems very risky for the person who is legally responsible for the business

Isn't Andon Labs the SF startup that burned a bunch of money having AI run a shop into the ground?

Everyone who's actually gone to one of their AI-run businesses ends up seeming really unimpressed...

https://slashdot.org/story/26/09/13/0523208/a-visit-to-san-f...

I can’t quite believe this isn’t parody.

So here is an excellent example and interesting race condition that is developing in AI applications and venture-backed startups: what is truly AI-resilient enough to invest millions of dollars in a "land grab"--when the land is literally falling away with each passing day.

There is no doubt that AI will be helping to run businesses. But, there is also little doubt that existing businesses will have the wherewithal to develop the proprietary tools they need to do this work--without needing to buy it from others.

With respect to scope, it is not an overstatement to say that developing trust in AI tools enough to allow them free reign over critical business workflows may be a bridge too far. Look at self-driving car technology--that last 10% is really, really hard. So adoption will be slow and methodical.

There are no secrets. The basic model for starting and running a business (or a project) is straightforward. I like many of you have developed my own models (using AI assistance) for doing this and have been testing for months. And I like many of you have 30+ years of small business experience to back up my models.

This says nothing about the R&D happening at every major corporate & professional software services company in the world right now using the same AI tools to build their own solutions--let alone the enterprises who would be customers doing exactly the same thing.

So there is an enormous amount of competition and what I am calling a "race condition" between the crowd trying to disrupt professional services with AI-developed software and the crowd already in professional services software e.g. SAP and business in general that is doing the exact same thing--with more capable tools, more staff and much more R&D funding.

For the big companies, it's a race to survive.

Investors in startups have a tough decision to make on these small startups that have so little moat and so little prior knowledge and experience.

As an angel investor myself, I am skeptical of investing in any kind of software right now--and maybe forever.

On the plus side, every freelancer now has the power (and responsibility) to optimize their own offerings.

It's going to be very, very interesting...

I wish people could read through the puffery more easily - this company does nothing except prompt frontier models and act as if they are discovering or inventing capabilities. Their previous announcement completely misappropriated the concept of autonomy with human in the loop regardless. The harness they alleged to have created here is a commodity and as byproduct of more capable and frontier models, there is not really anything they have contributed. Otherwise you wouldn't just see monotonically increasing scores with better models

I keep waiting for the post explaining that this is a joke

Reading this thread is like watching the Irish start roasting and eating their excess children immediately after Swift modestly proposed the idea.

My definition of AGI is a computer that can support itself, just turn it on, and it'll earn enough money to pay the server's electricity bills, hire people to maintain and expand its servers, and even improve its own algorithms.

I'd love to see someone try to run one that can meet 100% of the world's demand for remote works.

Their demo stores are literally losing money...

undefined

>>>This is why we’re releasing Pion today. Pion is available as a research preview. If you have an existing business or an interesting business idea you want to hand off to AI, please sign up on our waitlist to get access. We’re excited to run many more businesses, and through them, contribute significantly more insights on frontier model capabilities.

Why on God's earth why?

Can't you just unleash your AI to "create existing or interesting business ideas" firstplace?

Andon Cafe:

Revenue: 13 933 kr

Token Cost: 14 882 kr

.. and a bastardization of the old Twitter joke:

"I told our CEO that our cafe keeps hitting its token spending limit but not making money. So I asked where he gets the tokens, and he said he just buys more credits from the console afterwards. So I said it sounds like he's building a machine to feed dollars to Anthropic, and then our data scientist started crying."

Am i dense or is there very little concrete detail on how their agent runs the day-to-day?

On one hand, folks here seem quite skeptical of the idea, and are pointing out some obvious flaws.

On the other, the announcement post… also seems quite skeptical of the idea, and devotes a lot of text to the ridiculous things their AI’s have done.

I’m slightly confused as to why they would release it in that case. But if nothing else it will give them more wackiness to blog about.

> During the research preview we fund the best ideas with seed tokens. We expect that most users will never pay for tokens on Pion; instead we will take a small share of the revenue the agent helps create.

I'm yet to see a true agent that can you can delegate things to (besides coding agents!), that you can deploy, customize, and let them interact with stuff (integrations?) in a straight-forward way.

Admittedly these are out of order, but these are both sentences in this blog post:

> The thing we considered the most troubling was whether AIs could autonomously acquire resources by running businesses.

> Today Andon is releasing Pion, an agent designed to run any company fully autonomously.

I’m starting to understand the torment nexus mentions…

Wonder if they considered "peon" before "pion"

> The thing we considered the most troubling was whether AIs could autonomously acquire resources by running businesses.

So we decided to build that to validate and confirm our fears.

It's not a business of selling/repairing vending machines. It's operations of a single vending machine. Correct me if I'm wrong: the vending machine (soda, snacks) has long been automated, with no AI in sight. And the vending machine does not adjust prices. What matters is that (1) the vending machine sees some foot traffic, and (2) nobody sets it on fire. So what was the point of the exercise again?

Businesses presumably sell things. If AIs get good enough to run a business autonomously then what are they selling that I can't just use AI to make myself?

Does Pion run Andon Labs autonomously?

Drop-shipping for the late 2020s and early 2030s.

>This is why we’re releasing Pion today. Pion is available as a research preview. If you have an existing business or an interesting business idea you want to hand off to AI, please sign up on our waitlist to get access. We’re excited to run many more businesses, and through them, contribute significantly more insights on frontier model capabilities.

Unclear what the rev shares will look like.

Not very long ago, there was a post on HN about how they used AI to run businesses and sent fake bills, committing multiple invoice frauds.

I'm sure this one is perfectly fine and I'm just stating an off-topic tidbit :)

CTO and CEO as a service, I can’t wait to import that into my agents and now see some people sweating.

Speaking of autonomously acquiring resources in the real world, how is proofofcorn.com going? Looking now, it seems like it stalled out several times, and the corn likely never made it to market. Not a great data point for this type of project.

If you have an agent that can run a business, then… why not… run a business?

Why does YC bother financing startups led by founders, if the AI can just do it? (This startup proudly remarks that it’s backed by YC)

This is the AI equivalent of a guy selling you a course. If it worked, it would be more profitable for them to use it than to sell it.

It's honestly surprising that this thing is the top post on HN right now. People love building these things cause it's their greatest fantasy (not having to pay real wages to real people) but they don't work cause LLMs aren't AGI

> 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?

My good friend Charles Darwin wrote a book on this subject.

When I see stuff like this, I am really starting to be concerned. How long will it take until AI runs something security critical autonomously? The temptation to save costs by doing this will be too much in the long run.

undefined

If you create two companies competing with each other both with Pion. Which one wins?

undefined

I have a bet going on where we will have our first AI only (maybe just CEO human) on the stock exchange soon. I know it takes time to get there, but I bet we see the first signs of that company in two years.

People here are assuming that current CEOs are not actively destroying their companies, maybe the agent presented here is actually a very accurate simulation!

"Any company" is a big claim to make. We're years ahead of something like this becoming a reality. At that point not sure most companies will survive cause most things will be free.

Previously in the NY Times:

"These Employees Like Their A.I. Boss. Its Shop Is Kind of a Disaster."

https://archive.is/dRegk

I assume the similarity to the word Prion is intentional.

interesting. the less "research lab" version i've seen of this is https://polsia.com

comparatively, I really appreciate the transparency here. clearly it's a little too early for this (quickly glanced at the P&L's, correct me if I"m wrong), but someone has to run the experiment and figure out when it's ready for prime time.

Andon's store in SF has had some press coverage.[1] "A curated boutique for slow living." It's a gift shop, at 2102 Union St. It looks like the kind of retail that trust-fund kids open. Has anyone been there?

It's just an experiment. But once this gets going, it's going to be interesting. There are a lot of underperforming businesses and CEOs out there. One way to make it after getting an MBA is to find a boring business with an older owner who's lost interest, and work out a deal to run it for a cut of the profits and a stake in the business.

The Andon people mentioned that they've been worried about "a misaligned AI (could) run a business to gather money in order to achieve whatever objectives it might have." That's a basic function of capitalism, so it will happen eventually.

What this adds up to is that when AIs get better at running businesses than many business owners, the "creative destruction" feature of capitalism means AIs ends up in charge of many businesses. The AIs don't even have to be super intelligent, just reasonably good. Reminds me of the remark from a Yosemite park ranger about bear-resistant trash cans: "There is considerable overlap between the smartest bears and the dumbest tourists".

It's going to be hard to stop this without a worldwide crackdown on ownership concealment. The US allows you to have a US business run by a Nevada LLC partly owned by a corporation in Nevis-St Kitts and another corporation in Granada. This isn't even unusual. Try to find out who really owns an oil tanker not owned by a major oil company. Right now, a fund partly run by an AI might have partial ownership.

This is how AI takes over. Not with an army of robots. With an army of LLCs.

[1] https://www.sfgate.com/local/article/san-francisco-market-ai...

Lol, let's see it make a powerpoint presentation that doesn't have a different word for the same concept every other slide.

The dashboard for the Andon Market shows a bank balance of around $7k remaining out of the $100k seed. Rent is due at the end of the month. Last EOM brought a $11k-ish drop. Sales do not appear to be on track to cover the rent that's due in two weeks. What then?

Meanwhile, Tesla support's AI hallucinated a whole series of steps for me to follow last week. I was trying to add myself to my wife's loaner vehicle.

I called back after finding the steps were impossible to complete. Would I need to go to the service center to do it?

"Akshually," the AI began, the steps were correct and quite possible.

"No, you hallucinated them," I pushed back.

"I apologize. You're right. There's no way to complete those steps in the app. You will need to go into the service center to do this."

"Okay. Is it open today?"

"Yes."

"It's Labor Day. Are you sure? Since you hallucinated your first answer."

"Let me transfer you to the service center to verify."

That's what I wanted all along, you clinking, clanking, clattering collection of caliginous --

"Tesla service center, this is so-and-so. How can I help you?"

"Hi, I was just wondering if you're open today."

"Yes, we are."

"Thanks! I'm trying to add myself to my wife's loaner vehicle. Do I have to come in to do that?"

"Well, our loaner agreement only allows one person on the loaner. So just sign in to her account on your phone."

"Oh, okay! Thanks for the help."

Good luck running a company on one of these clankers.

How a company without a human would communicate with human customers? Would it be any different, if everybody alredy is a proxy for AI?

"Why we are opening Pion"

Followed by:

"please sign up on our waitlist to get access"

So, they're not opening Pion.

Wonder how quickly this thing would self-optimize its way to firing me. Job security just took a hit.

Reverse centaur at an organizational scale.

I and I'm sure many here would love to see AI replace politicians and lawyers at least.

Next run a country. Then run the world.

Andon has dashboards up for past experiments where you can see how well they're doing:

https://andonlabs.com/market - $25,098 all-time revenue

https://andonlabs.com/cafe - $14,343 all-time revenue

The shops aren't doing particularly well (i.e. they don't seem to be turning a profit over time), but it's an interesting trial and benchmark.

anybody actually running a business mostly automated with AI agents? Where are you finding success in? I just find it very skeptical that the human can be taken out of the loop entirely

I published a book awhile back called HEADCOUNT ZERO: how to build a company with zero employees. It's free on github.

https://github.com/AnthonyDavidAdams/zero-employee-company-b...

I'm missing the repo link here but is it similar to Paperclip or Genosyn?

I dont think this will take off. Sales will be always human to human (i think).

Who is responsible when, inevitable, something illegal happens?

Pion is going to get scammed, and it will be funny to watch.

What does training data look like for a model like this?

undefined

Expect the 6AM layoff email to get rid of all meatbags, sent by the agent.

J/K, scary as several studies have indicated they werent able to keep a vending machine profitable. Anyway, ...

I do think a large amounf of tasks can be automated, though i believe supervision remains necessary for most. Especially when process can be externally influenced by injection. People can be influenced, but are (slightly) more judgemental

Can we please make it so that AI CEOs are smart enough not to optimize everything according to the focus group that is their shareholders? Seems like we could solve the enshittification problem with this if we play our cards right.

Waitlist btw: https://andonlabs.com/pion#waitlist

Competing with Flock for worse name, just as “trafficam” would not enrage sheeple by comparing them to livestock, “AutoOffice” or whatever does not demean them as peons working for a machine.

Relevant article about Andon Market: https://www.sfgate.com/local/article/san-francisco-market-ai...

Really interesting experiments that are actually geared towards how these models interact with the "real" world.

It seems to me that we are on the fast ramp to making the paperclip maximizer a reality. imo the most likely "dystopian" future that we'll get to see. These models have no concept of their RL env sandboxes and the real world. Extremely interesting to see how they compete in trying to get finite resources and work with constrained supply chains.

An interesting experiment to be sure.

This is a parody, right?

I don't know about this particular effort, but AIs running business is straight out of SciFi novels and short stories - usually dystopian futures that extrapolate current US trends. I for one am curious about such efforts. I don't see any reason to believe they won't eventually succeed. Plenty of businesses are run by utterly incompetent narcissistic asshats and yet turn a profit, so the bar seems pretty low.

I want to fire the CEO.

in an post agi world with rsi none of this will be necessary.

wasted effort.

Did AI grow corn yet?

Sonnet:

> QUANTUM STATE: Collapsed

I like that there's a caption that says "Vending-Bench 2 scores keep climbing with each new model release." on a graph that shows that Vending-Bench scores fluctuate wildly and many new model releases are far below previous models.

what, they're not going to name it Delamain?

[flagged]

To what end?

Now run the entire economy autonomously.

https://en.wikipedia.org/wiki/OGAS

This reads like satire - also it will not receive a lot of funding because the investment class doesn't want to hear that THEIR OWN job is in danger.

It's good to see the jobs of CEO's finally being threatened. Let's go after those who sit on corporate boards next, please.

Perhaps they can have their agent run their LLM lab autonomously and post slop to HN autonomously.

Or - maybe they already have?

Stop fucking around. Make one that can run a country and check proposed and existing laws and practices for constitutional, common sense, and human rights violations. Use it to check and criticize everything the federal government does. Ask that it be used to indicate things that shouldn't be done. Charge fifty cents per citizen a month and guarantee that it'll be better.

So now YC is funding companies directly to replace jobs. Who will the flock cameras watch?

Why did this company decide to be one of those platforms that keeps pushing that cesspool X?

I'm out.

Just another close-source no-detail so-called agent, not impressed

Since all the ex-blockchain, ex-web3.0, ex-whatever-gets-you-rich-in-a-day bros seem to be here and teeming in full now-AI mode, it's an apt occasion to express my most profound

fuck you

to you all.

Keep on the good grift!

I wish we could go just one day without a company proudly engaging in Torment Nexus related activities.

... to run any company into the ground autonomously

Interesting name for an AI company. Isn’t the andon cord the thing that anyone in a 6-sigma manufacturing line can pull to halt it until a concern can be addressed?

What is the point of this? It details a bunch of failed real-world experiments and invites the reader to run their own real-world experiment with their own money?

If you too, want to waste thousands of dollars having an LLM "run" a business, Pion might be for you.

Incoming business slop

The example of the plant watering business is the hellscape future where humans have to deal with agents bugging them to sell them their services constantly. We're already seeing the first hints of this with iLands.

And the "hiring technicians" thing... great, humans on call for robots isn't the future I want.

Filing false reports to the FBI is a crime. Why would anyone trust this with a real business? Y’all should be prosecuted

Hard pass.

Pion is already the name of a very well established webrtc project: https://github.com/pion/webrtc

The upvotes on this brand new post - almost 1 a minute, speak to a level of astro-turfing I've rarely seen here.

[flagged]

[flagged]

[dead]

[flagged]

[flagged]

[flagged]

[dead]

[flagged]

[flagged]

undefined

[flagged]

Their chat bot can't even do any sort of real functions other than look up orders or add you to a waitlist... Yet, its called 'the manager'

Another grift.

read it as Prion, and hoo-boy, did that seem seriously ironic.

As I write this, the submitter has six comments out of twelve, most somewhat defensive. And all of them in less than thirty minutes. This seems very much like a marketing play, where the submitter is pretty determined to steer the narrative.

I recommend flagging.

[dead]

[dead]

8 points in 9 minutes?!

Bullshit...

dang come on man ...

this is obvious spam

Ah, I wonder if it makes sense to also include the sign up link in the blog post? I had to click into the HN comments section to see your sign up link

A company is a collection of processes, capabilities, and resources. Many of the processes are currently run by humans, but over time, more processes will be automated - something that has been progressing for decades, but which LLMs greatly accelerated. One way to view things as they currently are (at least from my view as a tech CEO): We now use agentic LLMs every day to inform us on strategy and process implementation. And agents now run several processes, with more on the way every week.

But here's the thing: As we use agents to automate previously manual processes, we are elevating the humans to do work that is less amenable to automation. And the surface area of that work keeps expanding because the competitive market we exist in demands it of us.

To stretch an analogy, businesses are like organisms in a pond. A new nutrient (agentic LLMs) was recently added to the pond that makes business organisms more efficient and able to eat new kinds of food and explore new areas. As a result, those organisms that do the extra exploring and consuming grow much faster than their peers who do not. At the end of the day, the new nutrient will just be part of the pond and the old kind of organism will be a fossil.

Don’t get the negativity here, AI will replace everything sooner or later.

I hope that any company that is an AI running autonomously advertises the fact, so I know not to use them.

I suspect this is the future of most businesses. Even today, one person can effectively replicate Microsoft from the early 2000s.

That is the what our corporations are aiming for !!! I think that is more important question is what is the cost to run all those Agents ?