He was right, and it did not matter.

A backend engineer flagged a race condition in a planning meeting. He was right. It shipped anyway, took down checkout for two hours on a Friday, and in the post mortem somebody asked whether anyone had seen it coming.

He had. Here is what he said, nine days earlier, word for word off the recording:

I just think that maybe we should possibly look at the retry logic, if that makes sense? I might be wrong.

Six qualifiers wrapped around a correct technical warning. By the time the warning arrived, everyone in the room had been told how much weight to give it, and they gave it exactly that much.

The wrong thing to worry about

There is a whole industry telling you to stop saying um. It is aimed at the wrong target. Listeners are remarkably forgiving of disfluency, and there is decent evidence that a moderate rate of um and uh actually helps a listener by flagging that you are choosing your words carefully.

Nobody has ever lost an argument to an um. People lose arguments to the small pile of qualifiers they put in front of the thing they came to say.

What a hedge is really for

A hedge is a pre-emptive retreat. It builds a position you can fall back to if you get challenged, which is exactly why it feels safe. You never have to be wrong, because you never entirely said anything.

The trade is worse than it looks. The protection is real but small: nobody was going to humiliate the engineer for raising a race condition. The cost is that his idea now needed somebody to champion it, and he had just publicly declined the job.

You cannot hedge a claim and then be annoyed that nobody took it seriously. The hedge was the instruction.

What to cut, roughly in order

• Just. The highest frequency offender by a distance. "I just wanted to check", "just a quick thought". Its only job is making the next thing sound smaller than it is.
• I think, in front of something you actually know. Fine when you are genuinely unsure. Corrosive as a reflex.
• Maybe, possibly, perhaps, when stacked with another hedge. One is a softener. Two is a retreat.
• Sort of, kind of, a bit, attached to a firm noun. Something is either a problem or it is not.
• I might be wrong, but. You might be. So might everyone in the room. Announcing it first buys nothing.
• Does that make sense, used as a full stop. It sounds like checking for clarity. It works as asking for approval.

The same sentences, rebuilt

A word about upspeak

Rising intonation on a statement, turning a claim into a question with your voice, does something similar. It also attracts far more criticism than it deserves, most of it pointed at young women, so it is worth being accurate rather than joining in.

Rising intonation does mark uncertainty in most English varieties, so on a claim you are sure about it works against you. It is also a normal feature of several dialects and gets used constantly by confident speakers to check a listener is still with them. The melody is not the problem. Uncertainty you did not mean to express is the problem.

Why deciding to stop does not work

Hedges are not composed, they are automatic, and they arrive fastest under exactly the social pressure where you most wanted to be clear. They are also inaudible to the person producing them. To the engineer it had not sounded like hedging. It had sounded like being reasonable, and like leaving room for people who knew the system better than he did.

Which is why this is one of the few speaking habits that responds properly to counting. He highlighted every qualifier in one forty minute transcript and got to thirty one. What actually landed was not the total. It was noticing that they clustered, almost perfectly, in front of the four or five sentences he had cared most about.

That is the pattern worth looking for in your own. The hedges are not spread evenly. They gather around the things you were least willing to be wrong about, which is to say the things you most needed somebody to hear.