Introduction #
The 12 weeks have passed and I have finished the chip timed 5k race. In those 12 weeks I did not have perfect attendance and missed about 2 weeks of workouts (1 for being sick, 1 for being on vacation). Aside from these misses, I stuck to the AI based workout routine for the duration of the experiment.
Metrics #
Body #
For the first metric I have my weight which went down by 8 lbs. over the 12 weeks. This was less than the 1-1.5 lbs. a week recommended by Grok, but given my lax approach to diet and the skipped two weeks, I think the performance was adequate. My scale (see part 1 for details) also gives a body age, which tracked somewhat closely with the decreasing weight measurements - giving me 4 years better in body age by the end of the plan.
For muscle and body fat, the scale did not give a massive difference between the start of the routine and the end. If I’m being generous, it appears there is 0.5 - 1% improvement in muscle and reduction in body fat over the 12 weeks.
Athletics #
For the performance goal of running a 5k in 28 minutes or less, this goal was clearly achieved by week 8 of the routine. Using Strava to track some of my runs I had a 32:23 5k time towards the beginning of the routine (September 6th, 2025) but by week 8 I was able to hit 27:01. My best time was the last run of week 12 with a PR (personal record) set at 24:47.
The official chip timed race that I ran was about 1.5 weeks after the official end of the workout routine and was clocked at 27:10.
Conclusion #
Overall, I am very happy with the AI-based approach to coming up with a training regimen. It helps average out the advice online and avoids bringing up every competing fitness theory that is bound to pop up in social media advice threads. The results were outstanding for a performance-based target and acceptable for a weight-based target. If I had followed a strict diet, I do not doubt that the AI would’ve been closer to predicting my performance over the 12 weeks.
