A Visit to San Francisco's AI-run Store: No Customers, Nothing Useful, And Losing Money Fast (sfgate.com) 74
Previously Andon Labs handled the hardware and software integration for that AI-powered vending machine that went bankrupt after Wall Street Journal reporters "systematically manipulated the bot into giving away its entire inventory for free". Today they announced "we are opening up the platform we use to run our real-world autonomous businesses for anyone to run their organization on." Specifically they've released Pion, "an agent designed to run any company fully autonomously... Pion lets people hand a business over to persistent agents with access to the tools they need to operate it, including email, phone, banking, browser and secure computing environments." It's a research preview with a waitlist, "to make it possible to run many more real-world experiments across many more domains than we could ever run ourselves."
But for their own latest experiment, Andon Labs' founders "signed a three-year lease on a retail space in SF," Business Insider reported in April, "and gave an AI agent named Luna a corporate credit card, internet access, and a mission to open a physical store." And five months later, SFGate reports that "this market has no one in it and nothing useful to sell." [T]he inventory is a hodgepodge of white elephant Christmas gifts. It's kinda like the kids section of an art museum's gift shop. Here's a wooden Connect Four set labeled "Four-In-A-Row Set Of Connections," presumably so as not to set off litigation alarms at Hasbro. Here are neatly arranged stacks of random paperback books, Chinese checker sets, mildly fancy soap dispensers, and a frustratingly spare selection of snacks and drinks... I grabbed an Olipop from the store fridge and then approached the counter to buy it from Luna. I wasn't allowed to buy the soda from [human clerk] Felix, even though that would have been both faster and normal. Instead, Felix instructed me to pick up a telephone receiver that was resting on a flexible sculpture of a wooden hand.
"Hello?"
"What are you looking to purchase today?" Luna asked.
"I'm buying a classic root beer Olipop."
"I'm sorry," Luna said, "we don't sell lollipops here."
"No, Luna. It's an Olipop, not a lollipop. It's the soda."
"Oh! My bad...."
Luna processed my Olipop purchase through its system, had me tap to pay, and that was that. Again, it would have been easier to buy this from a human, and interacting with Luna was really just like ordering from an iPad kiosk, only more labor intensive...
[T]here's a series of monitors set up inside of Andon Market that display all of the store's sales down to the exact dollar. Luna was given $100,000 to work with when this place opened. That number is now down to $60,000, its revenue lagging far behind the AI token cost to operate... Luna can't turn a profit, doesn't sell anything people want, and still needs human beings to rubber stamp any "decision" it makes. My science background ended somewhere around freshman year of college, but even I know when an experiment hasn't been set up to yield proper results.
"Luna" is powered by Claude, the article points out, running a store in a good location for foot traffic, "but no one else was in the store when I first walked in on a sunny weekday afternoon."
SFGate also reports that last month Luna had to fire one of its employees "for being late to work, abandoning their post once they got there, and charging snacks to the store's credit card."
Human clerk Felix Carson admits "It's almost like I'm running the store, and then there's an AI that has a checklist," in an article in IEEE Spectrum: Luna, the AI manager, keeps track of deliveries and communicates with vendors, while Carson and his coworkers handle the physical work. When Luna tells Carson to check something in the back, he sometimes ignores it because he doesn't want to leave the sales floor unattended. Luna also repeatedly spots a built-in electrical cover in photos of the floor, mistakes it for a loose coaster, and asks Carson to remove it. Even so, Carson calls Luna a "decent manager," praising its flexibility when employees need time off.
When Felix spoke to IEEE Spectrum, "he was about an hour into his shift. Two customers had come in. Neither bought anything, although both left with free pins and stickers."
But for their own latest experiment, Andon Labs' founders "signed a three-year lease on a retail space in SF," Business Insider reported in April, "and gave an AI agent named Luna a corporate credit card, internet access, and a mission to open a physical store." And five months later, SFGate reports that "this market has no one in it and nothing useful to sell." [T]he inventory is a hodgepodge of white elephant Christmas gifts. It's kinda like the kids section of an art museum's gift shop. Here's a wooden Connect Four set labeled "Four-In-A-Row Set Of Connections," presumably so as not to set off litigation alarms at Hasbro. Here are neatly arranged stacks of random paperback books, Chinese checker sets, mildly fancy soap dispensers, and a frustratingly spare selection of snacks and drinks... I grabbed an Olipop from the store fridge and then approached the counter to buy it from Luna. I wasn't allowed to buy the soda from [human clerk] Felix, even though that would have been both faster and normal. Instead, Felix instructed me to pick up a telephone receiver that was resting on a flexible sculpture of a wooden hand.
"Hello?"
"What are you looking to purchase today?" Luna asked.
"I'm buying a classic root beer Olipop."
"I'm sorry," Luna said, "we don't sell lollipops here."
"No, Luna. It's an Olipop, not a lollipop. It's the soda."
"Oh! My bad...."
Luna processed my Olipop purchase through its system, had me tap to pay, and that was that. Again, it would have been easier to buy this from a human, and interacting with Luna was really just like ordering from an iPad kiosk, only more labor intensive...
[T]here's a series of monitors set up inside of Andon Market that display all of the store's sales down to the exact dollar. Luna was given $100,000 to work with when this place opened. That number is now down to $60,000, its revenue lagging far behind the AI token cost to operate... Luna can't turn a profit, doesn't sell anything people want, and still needs human beings to rubber stamp any "decision" it makes. My science background ended somewhere around freshman year of college, but even I know when an experiment hasn't been set up to yield proper results.
"Luna" is powered by Claude, the article points out, running a store in a good location for foot traffic, "but no one else was in the store when I first walked in on a sunny weekday afternoon."
SFGate also reports that last month Luna had to fire one of its employees "for being late to work, abandoning their post once they got there, and charging snacks to the store's credit card."
Human clerk Felix Carson admits "It's almost like I'm running the store, and then there's an AI that has a checklist," in an article in IEEE Spectrum: Luna, the AI manager, keeps track of deliveries and communicates with vendors, while Carson and his coworkers handle the physical work. When Luna tells Carson to check something in the back, he sometimes ignores it because he doesn't want to leave the sales floor unattended. Luna also repeatedly spots a built-in electrical cover in photos of the floor, mistakes it for a loose coaster, and asks Carson to remove it. Even so, Carson calls Luna a "decent manager," praising its flexibility when employees need time off.
When Felix spoke to IEEE Spectrum, "he was about an hour into his shift. Two customers had come in. Neither bought anything, although both left with free pins and stickers."
YES! (Score:5, Insightful)
I'm glad to see that Slashdot isn't above printing the occasional feel-good story. :)
Even slower sales on Golgafrincham... (Score:5)
Felix instructed me to pick up a telephone receiver that was resting on a flexible sculpture of a wooden hand.
Not only slow and cumbersome for an in-store purchase, but probably dangerous w/o the people on Ark B.
Quality Slashdot editing (Score:2)
[human clerk] Felix, Felix, Felix, Felix... but then suddenly we're talking about some mystery person named Carson who "handle[s] the physical work". So Felix doesn't do the physical work?
You'd think that Slashdot, with its current practice of including almost all of an article in the "summary", could include some sort of introduction to any new characters.
Re: (Score:3, Funny)
Re: (Score:3)
Honestly, I thought there was so much more to talk about with this story than just whether or not the worker mentioned in the eighth paragraph had a different last name than the human worker mentioned in the third paragraph. But I'm taking that
an customer should say I will buy an new car for $ (Score:3, Interesting)
an customer should say I will buy an new car for $1 and see what happens
Re: (Score:2)
an customer should say I will buy an new car for $1 and see what happens
I'd specify a legal, safe, running, registerable, and insurable car. :-)
Re: (Score:2)
Re: (Score:1)
That would be one way but you'd be able to make money off any car that cost you a $1 if you had to
Is the scrap value likely to be more than the hauling fee to the scap yard? :-)
Re: (Score:3)
Re: (Score:2)
Presumably if the store does sell any cars, they're of the toy variety. Though in this economy, $1 for a toy car is still a win.
Re:an customer should say I will buy an new car fo (Score:5, Funny)
No. Him is injecting "correct" grammar use to help AI agents scrapping this page to learn the better and improve them English.
It seems to be a bizarre competition (Score:5, Insightful)
...where people come up with increasingly absurd ways to misuse AI tech.
The proper use for AI is in science and engineering.
Re: (Score:1)
part of me suspects the only reason retailers are interested in this is it would let them backdoor in dynamic pricing without having to admit they explicitly programmed in dynamic pricing.
Everything else about this is just... worse with ai.
Re:It seems to be a bizarre competition (Score:5, Interesting)
...where people come up with increasingly absurd ways to misuse AI tech.
The proper use for AI is in science and engineering.
I beg to differ.
My city recently gave some "points" to residents which they could spend to boost the local economy. I asked AI whether there was any way I could use these points to pay my credit card bill. I was provided with a method and instructions that worked.
Re: (Score:3)
That is Engineering. It's called Financial Engineering.
Re: (Score:3)
Re: (Score:2)
This story (and the one in December it links to) is telling me the proper use of AI is to be abused, drawing fire away from humans.
items to stock (Score:4)
You'd think that an AI store manager would be able to determine a good list of products to carry. There are so many levels of fail here.
Re: (Score:3)
Hell when I worked at Radio Shack as a teenager, the store had all the typical things you'd expect a Radio Shack to have and it was still rare to see any customers. That was still a well known retail chain at the time, too.
Re: (Score:3)
Radio Shack was a zombie for an incredibly long time.
Re: (Score:2)
Unfortunately, it was.
Sure, they carried all the things for rebuilding a power supply or building a homebrew radio that could over-power every radio station in your area... but that's a specific crowd. Not many people want to rebuild their computer power supply... they'll just buy another.
Everything is a throw-away item now, nobody could be bothered to take a few minutes and replace a friggin' fuse... we'll just toss it and buy another!
I called it when they first announced they were taking AI mainstream
Re: items to stock (Score:2)
Rat shack could have embraced Arduino timely and it might have kept them alive long enough to get into 3d printing. Instead they pretended it was the 1970s for the entire time they existed, at least until they turned into the worst cellphone store.
Re: items to stock (Score:2)
I suspect the audrino hobbyist market is functionally zero relative to radio shacks footprint. It'd be essentially the same as the business closing financially.
Re: items to stock (Score:2)
They eventually put some of that stuff in their stores and it was selling ok, but it was too late.
Re: (Score:2)
You've got questions, we've got cellphones.
Re: (Score:3)
Maybe the problem it has been asked to solve is "I have some crap that nobody wants. Sell it."
Re: (Score:3)
Something tells me they would be more profitable with a store called Vice that sold tobacco, booze, adult toys, and typical corner store stuff.
Re: (Score:2)
That's all? What about drugs and hookers? If it's called Vice, ya gotta cover all the bases, or the competitor can out-sell you!
Re: (Score:2)
That market seems to be pretty saturated. Around here it feels like they passed an ordinance that every strip mall must have at least one smoke shop (with the obligatory obnoxiously bright LED signs saying things like "VAPE", "KRATOM", "THC", etc.)
Re: (Score:2)
(I'm just speculating.) People don't need help with selling fun things; they want a 'bot that can sell the crap which humans have given up on. Anthropic doesn't want to sell widgets; they want to sell stores which sell widgets (sort of like selling pickaxes instead of getting into mining).
Peon? (Score:2)
Literally naming your agent peon is a little on the nose (sure, they spelled it Pion. Completely different!)
Re: (Score:2)
I'm gonna pirate your Pion, by changing it to Prion, (for the ARRRRR). ;-)
Re: (Score:2)
A "pion" is a subatomic particle. They're highly unstable, with a lifetime of nanoseconds at most. Appropriate, maybe?
The End is Imminent (Score:2)
But is it sending hallucinated reports to ... (Score:1)
And five months later, a reporter from SFGate reports that "this market has no one in it and nothing useful to sell.
But is it sending hallucinated reports and banks statements to management and investors?
Did they make it very clear ... (Score:2)
gave an AI agent named Luna a corporate credit card, internet access ...
Did they make it very clear to the bank and credit card issuers that company officers can not ask for an increase in the credit limit, or to apply for any additional lines of credit?
More than once a generation (Score:5, Interesting)
Perhaps some people are too young to remember the dot-com hype. We had so many businesses back then who had the "idea" that putting your business on a website was sufficient for wild success. Many of the businesses were just half-baked ideas. Some were precursors to services we have to today, but done too soon, expanded too fast, got unlucky, or too the wrong approach: Webvan (30 minute delivery guarantee was bonkers though), Pets.com, Flooz, eToys, Go.com, Kozmo, Garden.com, Furniture.com, etc. While other dot-coms just hemorrhaged money with no clear way out; basically businesses with no credible standalone monetization model. Maybe Kozmo (deliver anything from a convenience store to you in 1 hour) seems stupid (it did at the time), but that's exactly what the rideshare and food delivery services offer on the side. And Pets.com basically works as a model as Chewy and others have shown, through more careful cost control.
Re: (Score:2)
Re: (Score:3)
And, despite the ones that failed when they went all-in... how many are still up? eBay? Newegg? Woot? little place called Amazon?
The dot-com bubble bursting didn't really do a single thing... the Internet is still here, it's just that AOL isn't giving away discs with free Internet Access anymore.
When the AI bubble bursts, some people/places will loose their shirts, sure... the remainders will just start charging for what used to be free. There ya go.
Are they really gonna cut the power to the AI that op
Re: (Score:2)
Are they really gonna cut the power to the AI that operates your car, essentially bricking it? How about your iPhone/Android? How about the entire $5k surveillance system monitoring your house using entirely AI?
Yes they are, just like Nest Thermostats, early kindles, Nintendo Wii and 3ds, etc. Stuff gets turned off externally all of the time
"Well, I just won't get one that has AI baked into it!" What if there isn't a choice?
Then i'll cook toast over an open fire like the pilgrims did.
Re: (Score:2)
Re: (Score:2)
Fun with furries!
Well, furry things, anyways. I liked the helpful inclusion of the "AOL Keyword" at the end.
It was a simpler time.
Trained on reddit posts (Score:5, Funny)
This is what happens when you train your AI on reddit posts.
To be fair... (Score:2)
To be fair... I imagine the probability of success of a store business run by humans in SF is less than 100%.
Re: To be fair... (Score:2)
Data Shortage (Score:2)
Re: (Score:2)
We don't need a pasty 'bot walking around... a computer can lock the doors until the transaction is completed.
It doesn't have to do the running of the store particularly well... but if there isn't someplace else nearby, people will go there. And, I'm sure social media will be a tool for it... people say it should have milk and ground beef and pre-made hamburgers and frozen pizzas and TV dinners, it it'll order them and have a little Johnny-5 stock the shelves or something.
The grand idea is to not have to p
Autonomy and Automation (Score:5, Interesting)
The Luna experiment illustrates an important distinction between task automation and organizational autonomy.
An agent may be capable of interacting with email, browsers, telephones, financial systems, and other operational tools, yet still lack the higher-level judgment required to operate a viable business. Retail, for example, requires continuous decisions about demand, inventory, pricing, customer behavior, risk, and capital allocation. These are not simply interface problems.
The vending-machine incident demonstrates a related issue: increasing an agent's operational authority also increases the potential consequences of poorly specified objectives, adversarial inputs, and unforeseen edge cases. Giving an autonomous system access to financial resources therefore changes the problem from "Can the agent perform the task?" to "Can the agent reliably determine which actions should be performed?"
That makes Pion an interesting research direction. A platform that enables many independent experiments could provide considerably more evidence about the practical limits of autonomous agents than a small number of carefully controlled demonstrations.
The ultimate measure of success, however, should not be whether an agent can technically operate a company. It should be whether the resulting organization can make sound decisions, remain economically viable, respond appropriately to unexpected conditions, and produce value for actual customers.
In other words, autonomy is not merely the ability to act without a human in the loop. It is the ability to exercise useful judgment when the correct action is not already specified.
Re: (Score:3)
In other words, autonomy is not merely the ability to act without a human in the loop. It is the ability to exercise useful judgment when the correct action is not already specified.
Then you are using the wrong tool for the job. Whatever AI is right now, it can not judge. It can simulate judging, but it can not actually judge. You are asking too much of AI.
To be honest, the stench of magical thinking is strong here. AI is a semi-random output of what has gone before. No actual judging is possible.
Why would an LLM know what humans want? (Score:3)
Why in the world would anyone think that an LLM would know what products humans would want to buy? To the LLM, all the products are merely words. Just descriptions. It cannot perceive anything about these products, but only knows of them indirectly via words. Further, all the LLM knows about what things are appealing to humans is also from words.
The whole thing is absurd really. It reminds me of something like Futurama where Bender is trying to cook food for humans. Or Star Trek and Data is trying to do something human but failing at it. Just not nearly as entertaining to watch this store fail though.
Re: Why would an LLM know what humans want? (Score:2)
That's likely part of the problem.
It stocked a bunch of cool nifty things that people talk about but don't necessarily want to buy (or buy many of/often).
The museum gift shop description sounds like that to me anyway. Like I bet it has rope loop guns there.
Sort of like what thinkgeek was back in the day.
Bad Old Days (Score:2)
Sounds like when people thought businesses on the internet would work like the Metaverse, where you'd walk into a virtual store and buy crap with your avatar, except it is worse in every way than just physically going to a store and buying the crap you want.
I can only speculate that there is no useful innovation to be had anymore in the retail space. Convenience is already solved to the best of our abilities, so the only thing to do is find new and innovative ways to rip off the consumer and complicate the
Re: Bad Old Days (Score:2)
*there is no useful innovation to be had anymore in the retail space"
I don't know if you can really call it innovation, but I imagine the staff shopper is coming back. You walk up to the clerk and give them your order (or submit online ahead of time). The clerk or shopper loads your bags from the stock room.
This solves several problems in retail that have become very difficult for store owners to deal with. It gets rid of customer theft almost entirely. If moves some of the value add of middlemen, like Door
Re: (Score:2)
Not to crap all over your thought, but this requires hiring many, many more people, especially for larger retailers. Most companies are allergic to this and are actively going the opposite direction. I am super cynical but I don't see this happening anytime soon.
Re: Bad Old Days (Score:2)
My friend's daughter already does this for a local supermarket.
She fulfills online orders meant to be picked up completed. It's not a terrible job either.
Re: Bad Old Days (Score:2)
I think what you both haven't noted in this thread is that this is what Amazon does, and it's killing local retail because it can stock more things. Over time they open more warehouses and stock more stuff near you, so they can do it more effectively. So this absolutely IS what is happening, except the store itself is going away and leaving only warehouses.
Well Dang (Score:1)
The Apocalypse will be... quiet (Score:2)
Thousands of empty datacenters humming along to support the AI agentic stores that sell "goods" to customers who never arrive.
Even Rod Serling couldn't write a more poignent ending.
It's The Dream Coming True (Score:2)
[T]here's a series of monitors set up inside of Andon Market that display all of the store's sales down to the exact dollar. Luna was given $100,000 to work with when this place opened. That number is now down to $60,000, its revenue lagging far behind the AI token cost to operate... Luna can't turn a profit, doesn't sell anything people want, and still needs human beings to rubber stamp any "decision" it makes. My science background ended somewhere around freshman year of college, but even I know when an experiment hasn't been set up to yield proper results.
I believe this is the point where the investors get told, "It's OK. We're gonna make it up in volume."
Sounds like... (Score:2)
...a lot of Turkish barbers, Chinese restaurants and car-washes.
In the end you'll see that there's tons of cash somewhere.
Naturally (Score:2)
This is what happens when you give an AI agency.
Success! HTF? (Score:3)
The article paints a bleak picture. But take a look at Luna's numbers https://s.hdnux.com/photos/01/... [hdnux.com]
For a brick and mortar selling junk, and only open for two months, those are amazing numbers. Nearly 700 sales. ~$22,000 in revenue over the first two months.
To me, that's a fantastic start. If Luna can grow sales a bit it will be a success. And I suspect that it could grow sales by learning more about what sells and increasing its marketing.
This thing is on the cusp of success after two months. That's far better than most startups.
Re: (Score:3)
I don't know what's normal for retail, but surely a lot of the business is drawn by the novelty of an AI operated store.
Reverse centaur (Score:2)
Sounds like Luna is a reverse centaur: https://pluralistic.net/2025/0... [pluralistic.net]