• Even frontier models give wrong answers to financial questions more than half the time

• The relational reflexivity of frontier models

• The moral placidity of LLMs (and some thoughts on frontier models and scholarship)

• Will models fine-tuned to particular domains start consistently outperforming frontier models? Encoding habitus into LLMs

• The declining potential for agency with LLMs

• Claude 4 Opus on being terrified by its own capacity for malfeasance

• The frontier models still hallucinate wildly for literature searches

• Are Anthropic lobotomising Claude in order to turn it into ChatGPT?