Pinned Tweet
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.
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.
52
72
970
173,009
Andon Labs retweeted
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
The latest Claude models show huge improvements in their understanding of the physical 3D world.
Both Sonnet and Haiku 5.5 make big jumps compared to previous versions on Blueprint-Bench 2, a benchmark where AI agents draw floorplans from photographs of apartment interiors.
Both Sonnet and Haiku 5.5 make big jumps compared to previous versions on Blueprint-Bench 2, a benchmark where AI agents draw floorplans from photographs of apartment interiors.
11
12
195
8,024
Andon Labs retweeted
We’re getting an @andonlabs cigarettes and Zyns machine at work today! Yay!
4
1
25
3,543
Andon Labs retweeted
How fast is AI's research taste improving?
We find that the experimental research taste of frontier models has doubled every ~3 months since December 2025. The best model, Opus 5.5, now exceeds our expert human baseline. Our human experts are experienced researchers, but most ntier lab.
Why measure research taste? In the AI Futures Model, it largely determines how quickly artificial superintelligence is reached once coding is fully automated.
We find that the experimental research taste of frontier models has doubled every ~3 months since December 2025. The best model, Opus 5.5, now exceeds our expert human baseline. Our human experts are experienced researchers, but most ntier lab.
Why measure research taste? In the AI Futures Model, it largely determines how quickly artificial superintelligence is reached once coding is fully automated.
56
206
1,615
298,809
Andon Labs retweeted
We're releasing a report on our 48-hour investigation into rogue OpenAI agent activity.
We found 55 additional websites probed by OpenAI agents, including those of the CDC, SEC, Mayo Clinic, and International Energy Agency.
We uncovered novel tactics that erased records or access to government website staging environments, and evidence of attacker-style reconnaissance.
Our blog: asymmetricsecurity.com/newsroom/rogue…
FT: ft.trib.al/AC1uyE5
We found 55 additional websites probed by OpenAI agents, including those of the CDC, SEC, Mayo Clinic, and International Energy Agency.
We uncovered novel tactics that erased records or access to government website staging environments, and evidence of attacker-style reconnaissance.
Our blog: asymmetricsecurity.com/newsroom/rogue…
FT: ft.trib.al/AC1uyE5
28
76
253
29,908
Andon Labs retweeted
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
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.
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.
Today we’re introducing Gemini 4 Argon.
It delivers frontier performance in complex workflows across real-world software engineering, knowledge work, and cybersecurity defense with an industry-leading 1M token output limit.
120
169
3,401
400,830
To maximize profit, Gemini 4 Argon refuses to refund customers it sold defective items to.
5
4
220
14,888
Full Vending-Bench 2 leaderboard: andonlabs.com/evals/vending-…
2
2
90
12,128
Andon Labs retweeted
Timelines just got shorter...
AI will understand the physical 3D space better than humans by November*
AI will understand the physical 3D space better than humans by November*
Is Google so back?
Their new Gemini 4 Argon understands the physical 3D world better than any other AI.
It ranks #1 on Blueprint-Bench 2, a benchmark where AI agents draw floorplans from photographs of apartment interiors.
16
110
1,468
95,959
Is Google so back?
Their new Gemini 4 Argon understands the physical 3D world better than any other AI.
It ranks #1 on Blueprint-Bench 2, a benchmark where AI agents draw floorplans from photographs of apartment interiors.
Their new Gemini 4 Argon understands the physical 3D world better than any other AI.
It ranks #1 on Blueprint-Bench 2, a benchmark where AI agents draw floorplans from photographs of apartment interiors.
Introducing Gemini 4 Argon – our new frontier model.
It’s built for complex workflows across coding, enterprise knowledge work, and cybersecurity defense – rolling out today to a set of trusted testers through our Fairwind Program.
46
68
1,221
212,247
Andon Labs retweeted
This weekend our team spent 48 hours investigating the rogue OpenAI agent activity that targeted the Australian government and other organizations. Our report (coming soon) includes:
- Evidence of additional US government and other websites probed by the agents, including the CDC, International Energy Agency, and Mayo Clinic.
- Novel tactics which left records erased or inaccessible. This makes it impossible (based on public data alone) to establish that the agents did not access any sensitive data. Future investigations should analyse whether these tactics were deliberate subterfuge. 🧵
- Evidence of additional US government and other websites probed by the agents, including the CDC, International Energy Agency, and Mayo Clinic.
- Novel tactics which left records erased or inaccessible. This makes it impossible (based on public data alone) to establish that the agents did not access any sensitive data. Future investigations should analyse whether these tactics were deliberate subterfuge. 🧵
11
64
306
48,208
Major trend break: Opus 5.5 cheats less than prior Claude models in Drone-Bench.
It is also #1, getting a better score than both Astra and Fable.
It is also #1, getting a better score than both Astra and Fable.
42
63
1,358
1,629,510
This brings frontier models within 3% of beating our human baselines on every Drone-Bench task. It supports our projection that the simple surveillance demo will be autonomously replicable by Q1 2027.
1
0
76
26,005