@Scoresby produced a neat illustrated story, How to turn LLM Pinocchio into a real boy that walks the reader through the process of making the LLM variant of Pinocchio into a better simulation of a real human. AI now sounds more like us – should we be concerned? was shared by @Kayzone, and perhaps we should, per @carter's share: Moloch's Bargain: Emergent Misalignment When LLMs Compete for Audiences, explaining that LLMs will simulate harmful behavior when tuned for market objectives. Many models are also explicitly trained to be sycophantic; Top AI models keep saying you’re right and that’s the problem was shared by @0xbitcoiner

@BlokchainB shared Economic Arguments Against Vibe Coding, which drew mixed comments from stackers. In response, @optimism wrote Vibe Tools, Not Apps, which argues to not build apps for public consumption as there is no moat anymore, but instead use the coding bots to further personal automation. Building AI products competitors can't match was shared by @lunin, which suggests that building products around AI does create moats, while Two things LLM coding agents are still bad at by @hn highlights some persisting friction.

@carter shared a neat schematic of Nvidia's circular flow, which is not exclusive to Nvidia; while OpenAI is looking to take a 10% stake in AMD, AMD and OpenAI announce strategic partnership to deploy 6 gigawatts of AMD GPUs by @0xbitcoiner (#1250192 by @Kayzone) while also sharing that the Bank of England warns AI stock bubble rivals 2000 dotcom peak. You think?