As large language models become a routine part of office life, a curious workflow is emerging: people ask AI to expand a handful of bullet points into polished reports or emails, only for someone else to ask AI to summarise them back into bullet points.

In this episode, Fraser, Nick and Peter ask whether this middle step still serves a purpose, or whether we're witnessing the beginning of the end for traditional written reports. The discussion explores how information is compressed and decompressed, why summaries inevitably lose nuance, and whether AI can sometimes add genuinely useful context rather than simply padding out prose. They also examine the role of language itself, asking whether future communication might move beyond text altogether towards highly personalised, AI-mediated experiences.

Along the way they discuss information theory, business communication, the limits of human attention, visual communication, the relationship between language and thought, and some speculative visions of how humans might communicate with AI in the future.

• YouGov survey - 'How are Britons using AI at work?'
• Bit rate of spoken language
• Speed of human thought
• Bloom's 'Two Sigma' problem