The short letter
AI has made the long letter free. The short one still costs what it always did, and now the reader pays for the difference.
Insights and thoughts Andrew Wood
AI has made the long letter free. The short one still costs what it always did.
“I have made this one longer only because I did not have the leisure to make it shorter.” The line gets credited to Mark Twain, to Abraham Lincoln, even to Cicero. Blaise Pascal wrote it in 1657, in a postscript to the sixteenth of his Provincial Letters. So people have been apologising for long letters for nearly 370 years. The problem has outlived the telegraph, the typewriter and email, each of which made writing cheaper, and AI is only the latest and biggest jump. The line itself has been copied so many times that the source got lost along the way, which is a fair preview of what happens when words travel faster than anyone checks them.
For most of history, writing was the expensive end of the deal, and rewriting was worse, until the delete key came along. Now a two-page incident summary, a forty-point change plan or a beautifully formatted status report takes seconds to produce. Reading it takes exactly as long as it used to. The cost hasn’t gone away. It has moved from the writer to the reader, and in operations the reader is often a customer’s director deciding whether this needs to go to their board, or an engineer on a bridge at two in the morning.
That makes distilling a message more valuable, and much harder to fake. A summary keeps the shape of the original and makes it smaller. Distilling means deciding which part was the point, and you can’t do that for a problem you don’t understand yet.
Fixing things works the same way. In an incident I always ask for a working theory, even if it’s only a gut instinct. Put it as simply as you can, test it, and each test strips away something that isn’t the issue until you’re left with the thing that is. “Can you say the problem in one sentence?” is still one of the best questions I know, but nobody should expect the answer in the first ten minutes. On most bridges the one sentence arrives after a few rounds of guessing and checking, and a rough theory said out loud gets there faster than a careful silence.
This is where the old idea of T-shaped engineers, deep in one area and broad across the rest, deserves a second look. The usual case for breadth is that it lets an engineer do more. Now anyone can get a fluent, confident answer on almost any technology in seconds, so breadth matters for a different reason. It’s what lets you notice when the fluent answer is wrong. Depth tells you how to fix something. Breadth tells you whether you’re fixing the right thing.
Breadth used to be for doing. Now it’s for doubting as well. The engineers I’d most want on a bridge in a few years’ time are the ones who can read a polished AI summary of an outage and say, calmly, “that’s not the problem.”
I’m not exempt from any of this. I’ve sent long internal emails because I hadn’t yet done the thinking that would have made them short, and Pascal’s line is one I have to remind myself of more often than I’d like. As words get cheaper to produce, the people who stand out will be the ones who know which few the reader needed, and have the courage to send only those.
Pascal, at least, had the excuse of a quill.