I've been trying to use AI for knowledge work (course planning). Here's what seems to work so far: 1. Give each 'deliverable' a Karpathy-style LLM-wiki (for me, a deliverable is a course, or a talk. A sizeable piece of work, more than 1 day's labour) 2. Split the deliverable into 'sections' that can be worked on in independent threads during the day shift (more than one thread meads you can parallelize work in different agent threads, all reading from/writing to the wiki for shared…

Matt Pocock@mattpocockukSep 7

Knowledge work is SO much harder to automate with agents than code We underestimate how easy software development is for agents - Tons of high quality automated feedback loops (types, tests) - Well-organized free documentation, searchable on the web - Simple-to-use workspaces with version control (i.e. repositories) - Long history of systematising and delegating work (i.e. sprints, tickets, specs) Knowledge work has none of these advantages