Previously, I built a workout tracking system to let me record a workout on my phone and store it on my computer. Each time I used it for the first week or so I had ideas for an improvement and I addressed them all with a quick prompt or two. This really felt like the "aha" moment that proves the thesis of "Personal Software". Fixing a 10-second annoyance or making the UI very slightly better is seldom worth an hour or more of coding, especially for a quick side project, but now that time cost is next to nothing.
Workout Lookup
I intend to build some visualizations to determine workout progress and be able to further analyze where I can make improvements at the gym. In the meantime:
hey deepseek, give me a command to quickly look up previous exercises
actually, do a fuzzy search
there's a lot of results, add some UI to alternate background colors in the list to make it easier to read
show the max weight I've done at the end and highlight it
if I look up jogs, compute and display the MPH of the jog
> bun lookup bb incline press
...
2026-10-23 BB Incline Press 140x8, 150x8, 160x7, 160x6
2026-11-03 BB Incline Press 120x10, 120x10, 120x10, 120x10, 130x10
2026-11-09 BB Incline Press 130x10, 145x8, 145x8, 145x6, 145x6
2026-11-12 BB Incline Press 120x10, 130x10, 140x7, 130x10, 130x9
Max: 2025-09-29 BB Incline Press 90x9, 120x5, 140x3, 160x2, 165x1, 175x1, 182.5x1, 190x1, 195x1Pre-fill Next Workout
Remembering the full correct format of a workout, which is strictly parsed, was an issue. First, each of the 6 workouts I do per week is different, and I substitute a jog in whenever I feel like it. Every 2 weeks the number of sets and reps changes as well.
deepseek, give me a command to see the next workout based on what is last recorded in the DB:
> bun next
Week 6 · Period C · next: Legs A
assumed from last: 2026-11-13 Arms B (week 6)
Legs A
Block 1 — 4 exercises — 5x10 — no rest between exercises, rest 60s between rounds
BB Squat
DB Lying Leg Curl
DB Lunge
RDL
Block 2 — 4 exercises — 5x10 — no rest between exercises, rest 60s between rounds
Leg Press
Stand Leg Curl
DB Calf Raise 1L
Calf Presscan you pre-fill a draft email that I can just edit at the gym?
oh, also pre-fill the weights that I started with the last time I did this workout:
/* the draft email on my phone to fill out */
week: 6
routine: Legs A
BB Squat: 235x
DB Lying Leg Curl: 50x
DB Lunge 35x
...Turning random thoughts/occurrences into features
It occurred to me I might want to relate my weight to my workouts. I'm not really focused on losing weight, though when I do bother to keep track and weigh myself I do better than when I ignore it.
deepseek, give me a new command to track weight as well
I had a time constraint during one workout but don't want future graphs to show I regressed:
deepseek, add a notes field
I was reading about some workout measurements that compute something like "tonnage per hour":
deepseek, add a duration field
This one in particular makes me consider making an app/website to handle the input on my phone. Recording a workout via email has been fine since adding prefilled drafts, but if I want to track duration, an app with start/stop buttons would be convenient.
Final Thoughts
A few new features and a dozen enhancements that cost me only the time it took to write down a sentence or two describing the need. In the past, I still would've typed those exact sentences except they would have gone into a TODO file I'd never get around to. Now this is software I use every day and every time I notice the slightest bit of friction I can instantly improve it.
Matt Pocock's /grill-me skill (or /grill-with-docs) left the project in good shape for AI agents to handle. There were plenty of decisions at the beginning I would not have bothered to write down for "just a side project" but the ADRs the skills produced gave context on why certain decisions were made. Importantly, they all contain context on why other choices were not made. That negative context is really useful for keeping the AI from taking the project in wayward directions.
I've also been looking into visualizations and some advanced analysis of workouts and have been very impressed with the results. I expect this will take me more time to read through and process (multiple equations and studies have been cited from academic white papers), but I'm excited to dive into it!