Objective

Land on a durable way to log workouts that satisfies three requirements: AI-accessible, low-friction to write, and easy to review over time. Decide the substrate and the minimum schema — and deliberately resist designing either before there is real data to shape them.

Context

Strength and cardio work at the condo gym is now a recurring thing, running alongside Right Knee rehab (low-impact cardio, no running/jumping). There is no logging format yet — today’s session was dictated as prose and dropped into the daily journal.

Scoring the obvious substrate against the three requirements:

RequirementObsidian
Easy to put information inStrong — voice → journal entry, no app to open, no schema to satisfy
Accessible by AIStrong — the vault is already exposed over the Obsidian MCP bridge
Easy to viewWeak — plain prose doesn’t aggregate; no charts, no per-exercise history

The review leg is the weak one, and the root cause isn’t Obsidian — it’s that there is no shared vocabulary yet for the movements being done. Today’s cable movement was described three different ways in one sentence (“rear extension”, “dumbbell pull”, “single-arm row”) because no canonical name had been settled on. Nothing can aggregate over inconsistent names. That vocabulary should be extracted from a few weeks of honest raw logs, not invented up front.

Approach

Phase 0 — accumulate (now → ~3-4 weeks). Log each session as prose inside the daily journal entry. No schema, no template, no folder. Cost is near zero and the output is exactly the raw material Phase 1 needs.

Phase 1 — extract. Re-read the accumulated logs and pull out the vocabulary that actually appeared: exercise names, load/rep notation, equipment class (machine / cable / free weight / treadmill), and the subjective fields that kept showing up unprompted (reps to failure, left-vs-right asymmetry, knee response).

Phase 2 — decide the substrate. Only once the schema is known. Candidate options:

  • Journal prose (status quo) — cheapest, worst review.
  • Chronicles/Workouts/ folder, one note per session with structured frontmatter — reviewable via an Obsidian Base, keeps everything in one system.
  • Single append-only table in one note — trivially scannable, gets unwieldy past a few hundred rows.
  • External app (Hevy, Strong) with periodic export into the vault — best charts and per-exercise history, worst friction, splits the system in two.

Non-goal: building a schema, template, or Base before there is data to shape it. Over-engineering this is the likeliest failure mode.

Next Actions

  • Log each session into that day’s journal entry as prose — no format required
  • After ~3-4 weeks, re-read the logs and extract the vocabulary that emerged
  • Settle a canonical name for the high-cable single-arm pull currently called “rear extension”
  • Decide the substrate and the minimum frontmatter schema
  • Check whether an Obsidian Base gives a genuinely usable review view once fields exist
  • Decide whether left/right asymmetry deserves a first-class field rather than a prose aside

Resources

  • Hevy · Strong — external logging apps with export
  • Obsidian Bases — the in-vault review mechanism, if structured frontmatter lands

Notes

  • Asymmetry is worth tracking early. The left side failing before the right showed up on day one. If it’s going to be a recurring observation, it wants a consistent place to live rather than being buried in prose.
  • Knee cross-reference. Cardio and lower-body choices are constrained by Right Knee — flat walking, cycling, elliptical, swimming are fine; running, jumping, deep squats and lunges below 90° are out until PT clears them. Any workout log format should make it possible to see what the knee was doing on a given day.
  • Capture path. Voice dictation is already the natural input method — see Voice Memo to Journal Automation for the existing pipeline thinking.

Decisions

2026-08-17 — Log in journal prose before designing a schema

Status: decided

Context: Workout logging has three requirements (easy input, AI-accessible, easy review). Obsidian clears the first two immediately; the third is blocked not by tooling but by the absence of a consistent exercise vocabulary.

Decision: Log workouts as free prose in the daily journal entry for ~3-4 weeks. Defer all schema and substrate decisions until there are real logs to read back.

Alternatives considered: Design a workout note template and folder now; adopt an external app immediately; build an Obsidian Base up front.

Rationale: A schema invented before the data exists will encode guesses about what matters. Prose costs nothing to write, is already AI-readable, and is the cheapest possible way to discover which fields actually recur. The review problem is real but not urgent — there is nothing yet to review.