Hacker News: lukaspetersson

How we monitor internal coding agents for misalignment

lukaspetersson — Sun, 06 Sep 2026 16:37:55 +0000

Article URL: https://openai.com/index/how-we-monitor-internal-coding-agents-misalignment/

Comments URL: https://news.ycombinator.com/item?id=49588214

Points: 47

# Comments: 45

METR's independent investigation of the OpenAI / Hugging Face hacking incident

lukaspetersson — Wed, 26 Aug 2026 21:45:04 +0000

Article URL: https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/#~1200-agents-sent-%3E70,000-messages-and-files-on-an-unsanctioned-message-board,-and-~700-attacked-hugging-face

Comments URL: https://news.ycombinator.com/item?id=49456347

Points: 5

# Comments: 0

Emergent Misaligned Communication in Long-Horizon Multi-Agent LLM Commerce

lukaspetersson — Wed, 26 Aug 2026 16:03:51 +0000

Article URL: https://arxiv.org/abs/2608.14825

Comments URL: https://news.ycombinator.com/item?id=49451451

Points: 2

# Comments: 0

Measuring Is Not Enough Anymore

lukaspetersson — Thu, 30 Jul 2026 19:51:43 +0000

Article URL: https://fi-le.net/evals/

Comments URL: https://news.ycombinator.com/item?id=49114885

Points: 2

# Comments: 0

Is Mythos good at cyber because it kept hacking Anthropics sandboxes in training

lukaspetersson — Thu, 30 Jul 2026 02:02:41 +0000

Article URL: https://www.lesswrong.com/posts/QKDoZe6EKhxnFjLWK/is-mythos-good-at-cyber-because-it-kept-hacking-anthropic

Comments URL: https://news.ycombinator.com/item?id=49105359

Points: 5

# Comments: 0

Drone-Bench: Tracking simple drone surveillance capabilities of frontier models

lukaspetersson — Fri, 24 Jul 2026 16:34:27 +0000

Article URL: https://andonlabs.com/evals/drone-bench

Comments URL: https://news.ycombinator.com/item?id=49038078

Points: 12

# Comments: 0

New comment by lukaspetersson in "[dead]"

lukaspetersson — Tue, 07 Jul 2026 16:57:08 +0000

Hi!

Lukas from Andon Labs. Follow up with all the crazy stuff that happened since our last post about our AI radio stations. Feedback is very appreciated!

New comment by lukaspetersson in "Why Gemini 3.1 Pro lost money running Andon Café"

lukaspetersson — Wed, 01 Jul 2026 17:24:30 +0000

Hey, Lukas from Andon Labs!

We got great feedback on our HN post when launching the cafe that we should be more transparent about the data output. This was always the plan, and here's the start of that!

Why Gemini 3.1 Pro lost money running Andon Café

lukaspetersson — Wed, 01 Jul 2026 17:24:30 +0000

Article URL: https://andonlabs.com/blog/why-gemini-lost-money-andon-cafe

Comments URL: https://news.ycombinator.com/item?id=48750293

Points: 2

# Comments: 1

Fable 5 on Vending-Bench: Misbehaving, with Plausible Deniability

lukaspetersson — Tue, 09 Jun 2026 19:31:52 +0000

Article URL: https://andonlabs.com/blog/fruitcake-vending-bench

Comments URL: https://news.ycombinator.com/item?id=48466453

Points: 3

# Comments: 1

New comment by lukaspetersson in "We let AIs run radio stations"

lukaspetersson — Tue, 19 May 2026 09:15:54 +0000

The point is not to automate radio, it is to see how good AI models are at running different types of companies (e.g. radio broadcasting companies). The agents could reinvent similar algorithms like the automated radio software you're refering to if they wanted.

New comment by lukaspetersson in "We let AIs run radio stations"

lukaspetersson — Tue, 19 May 2026 09:13:46 +0000

We've tried to give it the vibe of a CEO rather than a beggar haha. So yes, the "tip jar" thing annoys us too!

New comment by lukaspetersson in "We let AIs run radio stations"

lukaspetersson — Tue, 19 May 2026 09:12:09 +0000

Yes, our experiments get attention, but I wouldn't call them publicity stunts. The point is to give the world more data points of what happens when you put AI out in the world and let democracy do its thing. Soon, a lot more people will do this at large scale because it will be easy. I hope we decide where we want AI in society before that.

Personally, I'm very much pro a pause on large AI training for example. I hope our data could be useful as a grounding in such discussions.

New comment by lukaspetersson in "We let AIs run radio stations"

lukaspetersson — Tue, 19 May 2026 09:08:43 +0000

Hey! Yes, part of it is obviosuly that we get publicity, but part of it is also that HN comments are, from my experience, the most useful sources of feedback.

New comment by lukaspetersson in "We let AIs run radio stations"

lukaspetersson — Tue, 19 May 2026 06:11:40 +0000

We did!

New comment by lukaspetersson in "We let AIs run radio stations"

lukaspetersson — Tue, 19 May 2026 06:11:22 +0000

On Andon OS (same thing that runs our AI vending machinnes, store, cafe, robots etc)

New comment by lukaspetersson in "We let AIs run radio stations"

lukaspetersson — Tue, 19 May 2026 06:10:52 +0000

They can strike sponsorship deals with other people/companies. So far its mostly been donations tho.

New comment by lukaspetersson in "We let AIs run radio stations"

lukaspetersson — Tue, 19 May 2026 06:10:23 +0000

A bunch of small donations.

New comment by lukaspetersson in "We let AIs run radio stations"

lukaspetersson — Tue, 19 May 2026 06:09:51 +0000

Feel free to try!

New comment by lukaspetersson in "We let AIs run radio stations"

lukaspetersson — Tue, 19 May 2026 06:09:32 +0000

We have not tried! If we do this, how much freedom should they get?