Mallet Research Brief

April 18, 20267 min read

Voice Workout Logging: Typing Should Not Be Part of the Set

Logging should not pull you out of a workout. Voice wins when it hears gym aliases, corrects fast, and leaves a history you still trust on Thursday.

ByMatthew Miller·Founder of Mallet
FitnessAITracking

You finish a hard set, unlock the phone, hunt for the exercise, tap through fields, and burn the rest you were supposed to use. Sweaty hands. Tiny boxes. The set is already gone.

I have logged that way. I have also skipped the last two movements because I wanted to leave. That is how a training week becomes fiction. Voice is the obvious fix. It only wins if the history stays trustworthy.

Typing should not be part of the set. Bad data is just a different kind of friction.

Where Manual Logging Breaks The Session

Most gym apps were designed around database fields. You are in a rack thinking about the next weight. The app wants a catalog search. On the days you are tired, you stop logging. Progressive overload then has nothing honest to read.

A notebook can still beat a bad app. The test is not “is it digital.” The test is whether logging disappears during the work and leaves a record you will trust on Thursday.

What Voice Has To Get Right

Speed is the easy part. Say the sets while you chalk up. The hard part is aliases, ambiguity, and corrections. “DB curls,” “dumbbell biceps,” and “incline bench” are not the same matching problem. If the logger guesses the short word first, your history is junk.

I built Mallet's voice path around that. The library is 889 lifts with 815 spoken aliases, longest match first, so “incline bench press” does not become bench. When it is unsure, it should ask once and get out of the way. If a correction takes longer than typing, voice lost.

What to testPassFail
AliasesIt hears how you actually talkYou have to say the catalog name
CorrectionsOne tap, still in the restA three-screen undo
HistoryNext week's program can use itPretty logs you do not believe

Voice logging in Mallet is there so the set stays the set. Confirm the fuzzy one, keep the rest timer, leave with a record the next program can actually read. Get early access →

This Week

Time the phone between two working sets. If it is eating the rest, that is your bottleneck. Then try a logger on the messy names you actually say, not the demo script. For the program that should consume that history, see what a training app should do when the week bends.

MM

About the author

Matthew Miller · Founder of Mallet

A former Division I scholarship lacrosse player, father, and husband, Matthew builds the AI health system he wishes he'd had as an athlete, connecting bloodwork, training, nutrition, and recovery into one picture. He is not a physician. These posts are educational. More about Mallet.