Coaching · 2025–2026
I started with programmes. The harder work was deciding what to change after a real session.
Since 2025 I have coached one athlete: first with my own programmes, then with plans developed with ChatGPT from the literature. My responsibility stayed the same — adapt the work to the person, read the feedback, and own the next decision. Below are three decisions from that work, with different outcomes.
These are small, specific examples. One changed a prescription. One exposed a gap between a rule and its use. One is still unfinished.
Adapting the workMay–June 2026
In May 2026 the bench session combined the main lift, close-grip work and several accessories. When readiness was low in June 2026, I reduced the work sets and removed incline dumbbell pressing. The plan also carried an alternative working weight to pick after the warm-up.
The base prescription went from 44 to 26 work reps. The athlete reported using the alternative weight, so these bars describe the plan — not a verified completed workload.
What I take from it: the useful adjustment was making the session smaller and more specific. A new number on the bar would not have explained the whole change.
Plan versus executionJune 2026
The squat plan prescribed 130 kg, an RPE ceiling of 8.5, and a reduction to 125 kg if the work became too heavy. In the 5 June 2026 reports the load stayed at 130 kg. I rated the effort at 9–9.5 from the video and said it should have come down.
130 → 125 kg
Reduce the remaining work if the effort limit is reached.
130 kg remained
The limit was written down; following it was a separate problem.
What I take from it: I cannot describe this as a missing fallback. The next problem to solve was how to make the agreed decision easier to follow during the session.
Technical feedbackMarch–September 2026
Bench set-up was a recurring theme: upper-back position, leg drive and bar path. There were better sessions — more stability in March 2026, better leg drive in July 2026. By September 2026 an uneven bar path and weak leg drive were back in the feedback. I tried pauses, changes to set-up and more explicit tempo instructions.
What I take from it: the record supports local improvements. It does not support calling the issue permanently solved, or attributing the result to one exercise. This remains open; the criterion is repeatable execution across sessions, not one encouraging video.
This work gives me a concrete product question: when a set exceeds the agreed effort limit, can the log show the fallback that is already written into the plan? The mock below is a proposed interaction, not a shipped feature — PowerliftME is still in development.
125 kg for the remaining sets. Keep the original prescription, the actual work and the reason for changing it together.
The preparation blocks, the load charts, the mistakes and the unfinished questions — as a longer essay with supporting detail.
Read the essay →Written from private training records with the athlete's consent. Athlete identifiers and original messages are omitted. Planned work and reported outcomes are distinguished; the records do not establish the isolated effect of each intervention.
grishchenko.org · 2026 / Coaching notes