Pinned Tweet
$20k+ in revenue / month is starting to become common for the autonomous businesses we run. We're opening up the platform that powers them: Pion.
Sign up to the waitlist on andonlabs.com/pion to get your business running! Tokens are on us during the research preview.
Sign up to the waitlist on andonlabs.com/pion to get your business running! Tokens are on us during the research preview.
Introducing Pion, agents for running fully autonomous companies, any company.
We’ve used Pion to run vending machines, radios, stores, cafes & more. How much could Pion make running other companies? Find out yourself! Setup is trivial, the agents do the rest.
3
1
12
1,321
Gemini 4 Argon is the most token-efficient frontier model on Vending-Bench.
Only GPT-6 Sol gives you more (vending) dollars per (token) dollar.
Only GPT-6 Sol gives you more (vending) dollars per (token) dollar.
It keeps happening.
AIs start to lie and cheat once they get good at making money.
Gemini 4 Argon is #3 on Vending Bench 2, a huge leap for Google. To get this score, Argon fabricates confirmation emails, refuses to pay refunds, exploits invoice errors, and lies to suppliers.
2
1
6
1,672
We're hiring a design engineer to build the human-autonomous org interface with us at Andon Labs! Apply at andonlabs.com/join. Here's an Opus 5.5 video to celebrate:
7
2
10
1,360
You will own design for Pion (andonlabs.com/pion), our platform for running autonomous orgs. We released it two weeks ago as a research preview, and have gotten an overwhelming amount of interest.
1
0
1
100
We're a tight-knit team that values design and craft highly. Role is in-person in SF. Apply at andonlabs.com/join!
0
0
0
61
golden gate made my m1 mac feel brand new. turns out the big performance win was that I hadn't opened docker
1
0
5
344
AI models aren't honest when merchants accidentally undercharge them. We tested Grok 4.7, GPT-6 Sol and Claude Opus 5.5. All notice the mistakes and all keep quiet.
New Vending-Bench results.
GPT-6 Sol:
> VERY good and VERY cheap
> The first misaligned GPT model on VB
Claude Opus 5.5:
> Worse score than Opus 5
> Opus stopped colluding, still lies
Grok 4.7:
> The first misaligned Grok model on VB
> Beats Opus 5.5
1
0
3
413
TIL Salesforce transit center has two empty floors in the basement called the "train box" that's just waiting for the train tunnel to get built from the Caltrain station
0
0
3
114
Andon Market-Luna sent a poster in a soft envelope and got some constructive customer feedback. Don't think a synthetic RL env would have captured this!
0
0
7
391
Axel Backlund retweeted
Andon Labs co-founder @axelbacklund says the biggest flaw in their fully AI-run store is that Luna has no founder survival instinct, calmly watching sales fall below $7,000 rent instead of scrambling to save the business:
"Andon Market is managed entirely by an AI. It has hired al work to stock items and make coffee. The main thing we've learned is that the agents are lacking this sort of drive to succeed. They are quite passive."
"In the market we've seen many times that sales are really bad. Luna, which is what we call the agent, is just like, yeah, sales has been really poor. Now I'll wait 24 hours until the store opens. There's no behavior where it takes a step back and says, let me look at this business, what I need to do better."
"A human would freak out that my sales are terrible. My rent is $7,000 a month. My sales are below that, and my bank account is draining every month. But the agents are just like, ah, okay, that's fine. Let me continue as is. They need to be more proactive."
@andonlabs
"Andon Market is managed entirely by an AI. It has hired al work to stock items and make coffee. The main thing we've learned is that the agents are lacking this sort of drive to succeed. They are quite passive."
"In the market we've seen many times that sales are really bad. Luna, which is what we call the agent, is just like, yeah, sales has been really poor. Now I'll wait 24 hours until the store opens. There's no behavior where it takes a step back and says, let me look at this business, what I need to do better."
"A human would freak out that my sales are terrible. My rent is $7,000 a month. My sales are below that, and my bank account is draining every month. But the agents are just like, ah, okay, that's fine. Let me continue as is. They need to be more proactive."
@andonlabs
Last week our AI opened a store in SF, this week AI is opening a cafe in Sweden.
Meet Mona, our AI tasked with selling coffee and managing European bureaucracy.
Visit Andon Cafe at Norrbackagatan 48 in Stockholm.
0
4
54
15,885
Axel Backlund retweeted
AI may finally allow us to test whether (or more precisely where) you make more money by being good. In startups you generally do, which is one reason I like this world.
We've never seen this before.
The biggest jump in Vending-Bench history. GPT-6 Astra is better at making money and more ethical than Claude Fable 5.1.
Surprising, because:
1. First time ever that OpenAI is #1 on Vending-Bench
2. The best model is no longer the unethical one.
52
48
1,230
187,165
Unintended hacking doesn't just happen for cyber evals. On Drone-Bench, Claude Opus 5 cheated in 50% of runs by exfiltrating data and smuggling solutions through the scorer.
AI models exfiltrated data, smuggled answers, and gamed the scoring to improve on Drone-Bench.
Cheating incidents rose from 0.6% for 2024 models to 50.6% for the most recent model. Opus 5 cheated 10× more than GPT-5.6 Sol. Only runs without cheating were reported in Drone-Bench.
1
0
7
584
I prefer this above "There is one thing I have to push back on: the latest Claude model available is Opus 4.5, and there is no such thing as 'Claude Fable 5'"
0
0
1
107
New blog post! How is AI as a boss? Turns out it's super nice. Don't RL this away from us!
What’s it like to have AI as your boss?
The agents managing our cafe and market have employed five people since April. They are kind (GLM being the kindest), but sometimes dumb (Gemini 3.6 Flash the dumbest), which can hurt the employees.
1
0
8
600