I have been using LLMs quite a lot lately, often when researching a new area for the first time. It's amazing how quickly you can get up to speed, with a few prompts. But when I came back to the topic, I would often find that my retained understanding was less good, with a sense that my learning had been somewhat superficial.
So rather than settling for a backlog of Claude conversations, I have started writing up my larger investigations into "memos", and publishing them. The rules are simple - while I lean heavily on the agent for research and data visualisation, the words have to come from me. After all, writing is thinking.
I've found this to be productive in terms of my own understanding, but also in terms of the quality of the research. If I'm writing something up, it forces me to confront strange findings and odd conclusions in an agent's output, which has often led to a clearer picture in the end. This will sometimes be a simple hallucination, but it will more often feel like an honest mistake interpreting ambiguous data. And it's only when I am writing that the possible mistake shows its face.
The topics are somewhat arcane (I have been focusing on Ethereum so far), which is why I chose to post them elsewhere. But you're welcome to take a look, and I'd love to hear what you think.