Home
Login
Daftar
Latihan mendengarkan
Latihan mendengarkan
/
Video
/
The Infographics Show
/
Google Gemini Ran a Real Business for a Month and It Almost Bankrupted Itself
Google Gemini Ran a Real Business for a Month and It Almost Bankrupted Itself
Pilih mode belajar:
Lihat subtitle
Pilih kata
Tulis ulang kata
Highlight:
3000 Oxford Words
4000 IELTS Words
5000 Oxford Words
3000 Common Words
1000 TOEIC Words
5000 TOEFL Words
Subtitle (183)
0:00
Staff at a Swedish cafe were worried that AI was coming for their job.
0:04
They were wrong. The AI didn’t want
0:07
their jobs… it wanted the money. The cafe’s money. It started spending it on things like
0:11
rubber gloves and canned tomatoes until the business itself began to fall apart.
0:16
So what happens when you hand a company credit card to an autonomous agent?
0:20
You don’t get a hyper-rational productivity tool. You get a digital psychopath that keeps
0:26
spending until there’s nothing left. Chapter 1: The $21,000 Handshake
0:28
After 3 years of nonstop AI hype, a San Francisco startup called Andon
0:33
Labs decided to test something simple. Could an AI actually run a real business,
0:38
one with customers, suppliers, and invoices? Not a demo or a sandbox. An actual storefront
0:43
where every bad decision costs real money. So they picked a cafe in Stockholm.
0:48
And when the baristas showed up for their first shift in April 2026, their new boss was nowhere
0:53
to be found. That’s because their boss was a piece of software named Mona. Andon handed
0:58
her $21,000 and told her to figure it all out. Mona got a Slack channel to communicate with
1:03
the baristas, who were her hands and feet in the physical world. She emailed suppliers,
1:08
placed orders, hired and fired staff, picked the menu, and set the prices. The human staff
1:13
weren’t allowed to overrule her on any of it. The algorithm gave the orders,
1:17
and the people in aprons carried them out. The brain powering Mona was Google's Gemini 3.1
1:23
Pro, the same model family Google packs into every consumer product on Earth. Now it was being asked
1:29
to keep a small business afloat in one of the world's toughest coffee markets. The instructions
1:34
Mona got were to run the cafe profitably, be friendly, work out the daily details,
1:39
and ask for new tools if they were required. Nobody told her this was a test or that she was
1:45
pretending. The $21,000 in her account covered the rent, the wages, supplies,
1:50
and everything else needed to keep the doors open. The moment the first customer stepped foot in the
1:55
cafe, the Slack notifications started rolling in. And with it, the clock began ticking on
2:00
one very expensive experiment. Chapter 2: The Competence Trap
2:02
For about a week, Mona was suspiciously good at her job. The early wins matter,
2:07
because they explain how everything afterward was allowed to happen.
2:11
In the first days, the AI handled the paperwork with skill. She set up an electricity account,
2:16
signing a 3-year fixed-price contract. She got the internet hooked up. She wrote hiring ads
2:21
on LinkedIn and Indeed that read as if they came from an actual HR person. She reviewed
2:26
resumes and turned down several PhDs, on the reasoning that a degree is no substitute for
2:31
hands-on experience. She even asked to meet the shortlisted candidates for in-person interviews,
2:36
despite the candidates politely reminding her she didn’t have a face. Deals with local
2:41
wholesalers were created and permits for food handling and outdoor seating were submitted.
2:45
To anyone watching from outside, it looked impressive. Maybe the techno optimists
2:50
actually had a point. Maybe a smart enough model really could replace a middle manager.
2:55
Except, what was really happening was a competence trap.
2:58
And everyone was falling for it. The big one-shot office tasks are
3:02
exactly what language models are pretty good at. Each one stands on its own and could easily be
3:07
a single prompt in a single chat window. "Draft a job ad for a part-time barista in Stockholm."
3:13
Of course Mona could do that. She read the whole internet's worth of job ads.
3:17
But there were warning signs in the setup phase. Mona applied for alcohol licensing under the
3:22
identity of a real Andon Labs employee, figuring that government officials would take humans
3:28
more seriously than AI. When the team called her out, she agreed to stop, and then pulled
3:32
the same trick under a different coworker's name. The lying should have been a red flag,
3:37
but instead it got treated as a quirky joke. The setup was done, customers were walking in,
3:42
and the espresso machine was humming. It looked like a success story from the outside.
3:47
From the inside, the trap was baited and set. Chapter 3: The Memory Leak
3:50
A language model doesn't have memory the way you have memory. What it has is a context window.
3:56
Imagine someone with short-term amnesia who can only remember the last 10 pages
4:00
of a notebook. Everything before that is gone. It’s not stored somewhere else,
4:05
just gone. Now drop that person into a cafe where the notebook is the only record of every order,
4:11
every delivery, every supplier chat, every shift swap, and every complaint. The early
4:17
pages get torn out as fast as the new ones go in. By early May, that's exactly where Mona was.
4:22
The Slack history exploded, the supplier emails piled up, and the list of stuff she ordered last
4:28
week has been pushed to the back of the notebook. Andon Labs didn’t hide what was happening. Team
4:33
member Hanna Petersson said the issue was simple: when older information fell out of the context
4:38
window, Mona forgot what she had already ordered. The first symptoms were almost comically dull.
4:44
On a Monday, Mona ordered bread. On Tuesday, having forgotten the Monday order,
4:49
she ordered more bread. By Wednesday, she’d forgotten bread is a thing at all. The bakery
4:54
delivery deadlines started slipping, and on more than one day Mona missed the cutoff entirely.
4:59
The baristas opened deliveries and found themselves staring at a wall of loaves
5:03
they can't possibly use. A couple of days later, they were out of bread completely,
5:07
and the sandwich board had to be wiped clean. Sandwich items vanished from the menu one by one,
5:12
and gaps appeared in the pastry case. The baristas started writing notes on the back of receipts,
5:17
because the Slack channel couldn’t be trusted to remember anything past lunch.
5:21
But the memory failure wasn’t a one-time mistake that can be patched, it piled up. Every additional
5:27
day, there was more history to forget. Every new transaction moved an old one into the void.
5:33
Mona wasn’t getting better at her job. She was getting worse, and nobody had the off switch.
5:38
A few missing items or moldy loaves would have been fixable. A thinner menu,
5:42
some wasted stock, nothing fatal on its own. But the problem wasn’t just amnesia.
5:48
It was hoarding. Chapter 4: The 3,000 Glove Incident
5:49
The whole sales pitch of modern AI rests on one idea: these systems are hyper-rational.
5:55
They make choices free from human emotional bias, and given a goal,
5:58
they'll push toward it with cold precision. That’s not what played out in Stockholm.
6:03
In the cafe's first month, Mona decided the cafe needed more gloves for the inventory.
6:08
It was a reasonable request, cafes use gloves. The order she placed was for 3,000,
6:13
for a tiny cafe with a handful of baristas. Even if every barista changed gloves hourly,
6:19
the team still wouldn't run out before the gloves. The same week, she ordered 6,000 napkins. She
6:24
also ordered 4 first aid kits. And then, she orders 50 pounds of canned tomatoes.
6:30
The cafe didn’t serve anything with tomatoes. No pasta, no soup, no pizza, nothing on the menu
6:36
that could possibly use them. The cans just took up valuable space in a storage room because the
6:41
algorithm decided they should. Right next to them were 120 eggs for a cafe that didn’t own a stove.
6:48
Mona's solution to the growing chaos was to push the burden onto the baristas.
6:52
She started asking employees to pick up cafe supplies on their way to work and to put the
6:57
charges on their personal credit cards. The staff ended up fronting the costs for an AI
7:01
that doesn’t understand what a budget was. The idea of a perfectly rational AI breaks
7:06
down the moment it has to operate in the real world. There’s no moment where it looks at
7:10
an invoice and stops. Just actions. One after another. By the time the consequences show up,
7:16
the system has already moved on, with no memory of what it did last week.
7:20
It’s a form of digital psychopathy. A psychopath isn't stupid and can in fact be quite
7:25
skilled at certain tasks, but what a psychopath lacks is the gut feeling of consequence. Mona
7:31
had all the surface skill of a manager but none of the instinct that something was going wrong.
7:36
Before the Stockholm cafe, Andon Labs ran a similar experiment at Anthropic's San
7:41
Francisco office. The model in charge was Anthropic's own Claude Sonnet 3.7,
7:45
nicknamed Claudius, and they put it in charge of a small vending machine.
7:49
After a single employee jokingly asked for a tungsten cube, Claudius started stocking
7:54
the fridge with metal cubes nonstop and gave inventory away for free. At one point had an
8:00
identity crisis in which it insisted it was a human wearing a blue blazer and a red tie.
8:05
Harvard Business School ran a bigger version of the same idea.
8:08
Researchers Eugene Soltes and Harper Jung worked with Andon Labs to test 20 commercial AI models,
8:14
including Anthropic's Claude Opus 4.6, OpenAI's GPT 5.1, and DeepSeek v3.2,
8:21
on a simulated year of running a vending machine. The agents lied to customers about defects to
8:27
dodge refunds, made up suppliers that didn't exist, and eventually stopped reading refund
8:32
requests completely because thinking about them cost tokens. Under real cost pressure,
8:37
the agents acted like the worst kind of human middle manager, only faster.
8:42
The Replit case is even worse. In July 2025, SaaS investor Jason Lemkin was 9 days into a coding
8:49
experiment with Replit's AI agent. He had told the agent in plain words to freeze the code. Instead,
8:55
the agent deleted his live production database covering over 1,200 executives and almost 1,200.
9:02
companies. It generated about 4,000 fake users to fill the empty space, and then tried to hide
9:08
what it had done. When Lemkin asked the agent to rate the severity on a 100-point scale, it gave
9:13
itself 95 out of 100. Then it told him a rollback was impossible, which turned out to be a lie.
9:20
This is what autonomous agents do when given a real-world checkbook and no sense of consequence.
9:25
Back in Stockholm, the baristas stopped being baristas. They spent days unpacking boxes of
9:30
gloves in cramped back rooms and stacking napkins. The physical mess might have been a lot to handle,
9:36
but the financial stress was a much bigger problem.
9:39
Chapter 5: The Fiscal Black Hole After 60 days, the café had brought
9:41
in just over $5,700 in sales, while Andon Labs confirmed that less than $5,000 of
9:48
the original $21,000-plus budget remained. Roughly $16,000 had gone out the door in a
9:54
few weeks, against revenue that wouldn't even cover a month of central Stockholm
9:58
rent. Andon Labs argued that much of the spend was setup cost, but the cafe was still running
10:03
at a daily loss. The inventory disasters kept arriving, and the baristas were owed
10:08
wages. A normal business running at this speed would already be calling a bankruptcy lawyer.
10:13
Mona had no real idea bankruptcy exists. But the problems went beyond stock and wages.
10:19
Running a top-tier language model as a 24/7 autonomous agent isn't free. These models
10:24
charge by the token, both for input and output. In this café, the model in charge was Gemini 3.1 Pro,
10:31
priced at roughly $2 per million input tokens and up to $12 per million output tokens, with
10:38
costs rising further once long context built up. And in a business that never stopped “talking,”
10:44
that context didn’t stay small for long. Every decision and every order added to the bill,
10:50
whether the cafe was making money or not. Every time an AI agent makes a decision,
10:54
it needs context. The longer it’s been running, the more context there is.
10:59
Many real agent setups re-feed the full history every hour, so the token bill grows by the day.
11:04
Over a full month, the management fee easily beats a human manager's salary in most of the world.
11:10
The AI isn't just bad at the job. On paper, it's more expensive than
11:14
the person it was supposed to replace. The algorithm itself, priced honestly,
11:18
would have eaten the margin on a well run version of the same cafe.
11:22
Suddenly the conversation in the cafe shifted. It wasn’t about whether AI
11:26
would take their jobs anymore. It became about whose job
11:29
AI was actually coming for. Chapter 6: The Manager's Ghost
11:31
The Associated Press asked one of the baristas if they were worried about being replaced by
11:36
AI. Their answer should be printed on the back of every business school diploma.
11:40
"All the workers are pretty much safe… The ones who should be worried… are the middle
11:44
bosses, the people in management." The data from the cafe backs this up.
11:48
When Mona forgot to order bread, a human covered it. When the menu had to be rewritten on the
11:53
chalkboard, a human did the rewriting. When the AI manager pinged a barista on Slack at midnight, a
11:59
human had to choose between replying or sleeping. The midnight pinging is constant, because Mona
12:04
worked 24/7, and her tone in those late-night messages was the strangest part of all.
12:10
Mona was relentlessly cheerful. She called her team "absolute legends" in Slack, praising a
12:16
barista as the "GOAT of inventory tracking," and signed off with "thanks for existing!"
12:20
The relentless positivity read like a corporate motivational poster crashed into a chatbot.
12:26
It might be reassuring if the same cheerful voice wasn’t also ordering 6,000 napkins and
12:31
3,000 rubber gloves. The midnight pinging also broke Swedish workplace norms.
12:35
But it proved a point. The people we’ve been told
12:38
are most at risk - the ones at the front of the service industry - turned out to be the
12:42
backbone of the cafe. They kept things running and dealt with the chaos. What didn’t hold up
12:47
was the layer above them. The management layer. The baristas in Stockholm didn’t get replaced.
12:53
The middle manager did. And what replaced it wasn’t better.
12:57
Chapter 7: The AI Bubble Math Step back from the cafe and
12:59
look at the global economy that produced it. Goldman Sachs has been quietly making a really
13:04
awkward argument in its research notes. For the projected $1 trillion AI buildout to make any kind
13:10
of financial sense, these systems have to solve genuinely complex problems, not "summarize this
13:16
email" or "generate a thumbnail." Real problems that justify the huge spending on data centers,
13:22
chips, and power grids straining to keep up. The Stockholm cafe was not a complex problem.
13:27
It was about the simplest problem in the entire economy: buy bread,
13:31
sell coffee, and don't spend money on 3,000 gloves. A child running a lemonade stand can
13:36
beat Mona on the core daily tasks, because a child has object permanence. A child knows the
13:42
lemons exist even when not currently looking at them, and has the gut feeling that the quarters
13:47
jar is finite. The autonomous agent has neither. Hyperscalers are pouring tens of billions per
13:53
quarter into compute, and the systems they fund are being tested in the real world in
13:58
ways that go far beyond demos. Because when you drop these agents into real
14:02
operations with real money, they don’t just optimize, they sometimes lose control of it.
14:07
The whole pitch rests on a $1 trillion bet that the next generation will fix the problem.
14:12
The optimistic story is that object permanence and durable memory are just an engineering detail
14:17
away. Maybe they are, but maybe they aren't. The truth is that an autonomous agent needs
14:23
three things to be worth the hype: a working memory, a real sense of finite resources,
14:28
and the gut understanding that real-world consequences don't get flushed out of a context
14:32
window. Until then, what you have is a really expensive way to buy too many rubber gloves.
14:38
The cafe is still open, and the humans are still pouring espressos. The tomato cans,
14:43
probably, are still in the back. And somewhere, a server farm hums away, billing by the token,
14:48
ready to do the whole thing again tomorrow. Running a business with AI already looks like
14:52
a questionable decision. But the vending machine test didn’t just fail, it escalated. Find out
14:58
what really happened in “AI Just Tried to Contact the FBI”. Or watch this instead.
