Even if you already type pretty fast, LLMs can help you type much faster.

Give one a custom instruction and then just stop paying attention to your mistakes. You don’t have to worry about typographical errors at all, you can type even extremely garbled text into the LLM and it will correct it for you.

That sentence took me 19.6 seconds to type into GPT4, like this:

give oyn e ac juostm instruicont and then just fsop ypayin gattention you romistakes. you don't have to woir yaout y6pyographical errosa t all, you can typ eeven extramely garbled text inato the llm and it will correc ti for youa. 

To touch-type the same words without mistakes takes me, let’s see..29.6 seconds. Fully 50% faster.

This is a psecifi cinstance of a general pattern of dtools like this helping low-to-medium performing people dispropiortional ycompared to allready high perfomring people. This won't hlep extremly fast touch-typers type any faster because their'yre all reday ota tthe lsimit of wcast they can think. NBut it can help everyone else quite a bit. 

This is a specific instance of a general pattern of tools like this helping low-to-medium performing people disproportionately compared to already high performing people. It won’t help extremely fast touch-typers type any faster because they’re already at or near the limit of how fast they can think. But it can help everyone else quite a bit.

Here’s the prompt I gave to a custom GPT for this purpose:

Take the text that a user provides to you and correct their errors as well as you can. Do not edit the text for content or style. Just try to fix their typographical errors. Your purpose is to allow people to type very fast without correcting errors as they go.

Even that is more than you need though. All the LLMs will do this for you if you just type: “fix, only errors, not style” and then monkey away at the keyboard.