Graham writes against the illusion of getting it right. His proposal is that serious work is not about predicting better than everyone else, it is about building positions that survive being wrong, because being wrong is going to happen.
What I took from it
Margin of safety, moved over to estimating work. An estimate that only closes in the best case is not an estimate, it is a wish in a table. Every proposal now carries, in writing, the assumption that would make it false if it fails: the historical data is not where they say it is, the one person who knows the process never becomes available, the source system has no usable export. Writing that line is uncomfortable and it prevents the ugly conversation in month three.
The moody business partner, which is the book's best metaphor. The market quotes you a price every day and you decide whether to listen. Applied to my trade: a client who is enthusiastic in January and cold in March has not changed problem, they have changed mood. Decisions get made about the problem.
And the distinction between the person who will put work into something and the person who will not, which Graham poses for investing and applies just as well to a project: choose deliberately how much of your own effort goes in, rather than pretending it will be a lot.
What I left out
The specific stock-selection formulas, which belong to another market and another era. And something admirers tend to skip: much of what makes the book readable today is the commentary added by the journalist who updated the edition, not the original text, which is dry and full of companies that no longer exist.