What we talk about when we talk about sideloading

Using AI to negotiate a $195k hospital bill down to $33k

EuroLLM: LLM made in Europe built to support all 24 official EU languages

Keep Android Open

Samsung makes ads on smart fridges official with upcoming software update

The AirPods Pro 3 flight problem

Boring is what we wanted

Tinkering is a way to acquire good taste

ChatGPT's Atlas: The Browser That's Anti-Web

We need a clearer framework for AI-assisted contributions to open source

Generative AI Image Editing Showdown

Ubiquiti SFP Wizard

Nvidia takes $1B stake in Nokia

Fil-C: A memory-safe C implementation

Hi HN! Our startup, Andon Labs, evaluates AI in the real world to measure capabilities and to see what can go wrong. For example, we previously made LLMs operate vending machines, and now we&#x27;re testing if they can control robots. There are two parts to this test:<p>1. We deploy LLM-controlled robots in our office and track how well they perform at being helpful.<p>2. We systematically test the robots on tasks in our office. We benchmark different LLMs against each other. You can read our paper &quot;Butter-Bench&quot; on arXiv: <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;pdf&#x2F;2510.21860" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;pdf&#x2F;2510.21860</a><p>The link in the title above (<a href="https:&#x2F;&#x2F;andonlabs.com&#x2F;evals&#x2F;butter-bench">https:&#x2F;&#x2F;andonlabs.com&#x2F;evals&#x2F;butter-bench</a>) leads to a blog post + leaderboard comparing which LLM is the best at our robotic tasks.

HN