Here are this week's links:
- How often do LLMs snitch? Recreating Theo’s SnitchBench with LLM. You might remember last week's story about Claude trying to contact the authorities in certain circumstances. It turns out that it's not just Claude and other models, including Deepseek, that also try to reach out to the authorities if cornered in a particular way and exposed to extreme prompts. Not LLM companies will likely face a tricky challenge: will they optimize for this behaviour or ... away from it?
- Google's slow death has begun. I don't necessarily subscribe to the headline, but Michael Spencer's post Google I/O article does a great job of showcasing both Google's impressive tech releases and the already visible impact on the web and search. Well worth a read for the stats and analysis. I suspect Google may be fine without web search: Android, Chrome, and Google Cloud are huge anchors in the storm, but how much consumer mind share can it afford to lose from search? This is a paid article, but you should be able to get a free preview.
- AI eats the world. This year's "big presentation" by Benedict Evans focuses on the AI disruption wave. Benedict is generally more sceptical about use-cases than others (which is healthy), but still points to significant rising trends. The presentation seems a little shorter on answers than those from previous years. AI has clearly increased the uncertainty about future outcomes.
- Jony Ive says Rabbit and Humane made bad products. I'm actually pretty excited about what OpenAI hardware might look like (though maybe not about how locked down it could be). Still, did Jony Ive really need to ding Humane, and especially Rabbit? The biggest indicator yet that they were onto something, but early. If there's a form factor I'd guess for the new device, something close to the rabbit + AirPods would be what I'd be guessing. The Verge article includes Rabbit founder Jesse Lyu's response (classy).
- ‘One day I overheard my boss saying: just put it in ChatGPT’: the workers who lost their jobs to AI. This Guardian article has interviews with a set of workers who recently lost jobs in AI-related shifts at their companies. There are no stats as to how much this is happening, but the stories do cover the dynamics in certain industries. No matter how much we say "new jobs will also arise," it's important to recognize that these changes are really happening and there is a real impact. Creative skills roles that emerged with digital tools and distribution, especially, are being shut down again as humans are taken out of the loop. It's hard to imagine this won't accelerate, and it's critical to figure out how those roles can transform rather than just be replaced.
Wishing you a great weekend.
