I’m reading Wittgenstein’s Tractatus Logico-Philosophicus, and some lines knock me out for half an evening.

So I read it with an LLM on the side. I photograph a page from my couch and ask what one line means. Even Fable struggles with some of them. More than once I’ve asked it to explain the explanation πŸ˜„

Somewhere between two of those photos it hit me what this replaces. A question about one specific line used to mean going to a university, catching a professor in a good mood, and hoping the mood stretched as far as propositions 6.34–6.36 that day. Two hopes, stacked. Now the answer lands before I’ve shifted on the couch.

But the instant answer is the smaller half.

The tip that actually improved my reasoning: don’t ask for the explanation first. Explain the line yourself β€” your version, wrong parts included β€” then let the model correct you. An answer tells you what the line means. A correction shows you exactly where your own thinking bent.

A professor’s good mood used to be the price of being wrong in front of someone who knew better. Now it’s free, at any hour, on any couch.