Paying the wrist cost

I was forty-eight hours into a long stretch of writing — half code, half documentation, half email — when the familiar ache in my right wrist showed up. I have lived with it on and off for years. It is not a diagnosis; it is a budget. The budget runs down as the day progresses, and around six in the evening I run out of it. The keystrokes I should have written instead get postponed; the high-quality reply to my manager turns into a one-line acknowledgement and a promise to write the long version tomorrow.

I have tried the standard interventions. Ergonomic keyboard, vertical mouse, screen height, chair posture, stretches at the top of each hour. They all help. None of them changed the underlying fact that I type for a living and typing has a cost.

The thing that changed the budget — for me, after a week — was moving the prose half of my day to voice.

The two-list rule

I did not move everything to voice. Code stayed on the keyboard. Shell commands stayed on the keyboard. File paths, identifiers, brackets, semicolons — all on the keyboard, where they belong. What moved to voice was anything that reads like a sentence: commit messages, PR descriptions, design docs, code comments, Slack replies, email drafts, the AI prompts I write inside the editor.

A rough estimate: somewhere between forty and sixty percent of the keystrokes I write in a day are prose-shaped. Moving those to voice moved forty-to-sixty percent of the wrist budget. The numbers matched the feel by the end of week one.

What it did not solve

Voice did not solve the underlying RSI. It is not a treatment; I want to be clear about that. The conditions that produce the ache are the same conditions they have always been. What voice did was reduce the load that aggravates them. The ache still shows up; it shows up later in the day, and at a lower intensity, because the typing budget has been spent on the typing that actually needs the keyboard.

There is a long literature on RSI interventions and I am not a clinician. What I can say from one user-of-one is that the wrist budget at the end of a long writing day is the metric I notice most, and voice moved it.

The hotkey ergonomics

The other thing I want to flag is the hotkey itself. Push-to-talk means the hold-key fires dozens of times a day. The wrong hotkey is a small repetitive-strain problem of its own.

I settled on hold-Option-Space, on an ergonomic split keyboard, with the Option key under my left thumb. The thumb is the strongest finger; pressing a key with it does not have the same load as pressing the same key with a pinky or with a finger I would otherwise be using to type. If you are setting up a dictation tool, spend ten minutes on the hotkey. Pick a combination that uses your strongest finger and that your editor does not bind. The downstream wrist budget over a year will reward the choice.

A more aggressive option I want to flag: a foot pedal mapped to the hotkey. For severe RSI cases, the wrists do nothing — the foot triggers the dictation. I do not need this; my partner, who has worse wrists than I do, has been experimenting with a USB foot pedal and reports it works.

What to expect after a week

If you are trying this, my honest expectations for week one are these.

You will say things you regret out loud at least once. The brain pace on dictation is faster than on typing, which is the point, but you have less of an internal review pass on what you said before it lands in a permanent record. Edit pass. Then send.

Your accuracy will rise sharply by Friday. The first day, the model will mis-hear half a dozen of your most-used names. By Friday — if you have seeded the dictionary as you went — the names that mattered will spell themselves the way you spell them.

You will keep typing the wrong things. Identifiers, brackets, shell commands. That is correct. The keyboard is the right tool for those. Resist the urge to dictate them; the time you save is smaller than the time you spend fixing what came out wrong.

Your wrists will be lighter by Saturday. Or you will not notice a difference, in which case the tool is not the limiting factor for you and the wrist health work is somewhere else. Either way, you will know in a week.

Picking the right thing for the right job

I want to end on the thing that I think is easiest to get wrong. Dictation is not the right tool for every keystroke. It is not the right tool even for every keystroke a programmer types. The argument is not “voice replaces keyboard”. The argument is “voice handles the prose-shaped half of the day so the keyboard handles the other half with budget left over”.

If you spend most of your day in code, voice will give you back the budget for the prose. If you spend most of your day in prose, voice will give you back the budget for the code. Either way, the metric that matters is whether your wrists still work on Saturday. Mine do, now. That is the part I wanted to write down.