No black box. This is exactly how your plan is built.

Your plan starts from a single fitness measure: VDOT, Jack Daniels and Jimmy Gilbert's pseudo-VO₂max derived from race performance. We take any recent race result you give us and convert it to an equivalent 5K using Pete Riegel's endurance model:

t₂ = t₁ × (d₂ / d₁)^1.06

The 1.06 exponent encodes how pace degrades with distance for trained runners. From the equivalent 5K we compute VDOT, and from VDOT every training pace in your plan. No recent result? We start from a deliberately conservative VDOT for your experience level — undertraining slightly beats overtraining injured, and you can rebuild the plan when you have a result.

Three pace zones do almost all the work in a race build:

Easy (~65–79% of VO₂max effort) builds aerobic infrastructure — capillaries, mitochondria, fat metabolism — at a stress level you can absorb daily. Most of your mileage lives here on purpose: volume you recover from beats intensity you don't.

Tempo (~88–92%) sits at your lactate threshold, the fastest pace your body clears lactate as fast as it produces it. Raising this ceiling is the single biggest lever on race pace for everything from 10K up.

Interval (~95–100%) develops VO₂max itself and the economy of running fast. It appears later in the plan, in controlled doses, because it's the highest return and the highest risk.

Plans run in phases, each with a job. Base establishes aerobic volume and durability — mostly easy running with strides. Build introduces threshold and interval work while volume keeps climbing. Peak holds your highest volume and race-specific intensity. Taper cuts volume sharply (roughly half) while keeping short sharpeners — fitness is banked; the taper spends fatigue.

Each phase's weekly pattern, intensity mix, and volume multiplier is authored per race distance — a 5K build sharpens with 800m repeats, a marathon build extends the long run and mid-week medium runs instead.

Weekly mileage scales to a peak fitted to you (your stated ceiling, or a level-scaled default), rising gradually inside each phase — the long-standing ~10% guideline. Every few weeks the plan deliberately drops volume 40%: a down week. Adaptation happens during recovery, not during stress; a plan with no valleys is a plan that assumes you never absorb training. Your long run is additionally capped at your stated longest run and at a fixed share of the week, whichever is lower.

With a race date, the plan's final week is race week. If your race is further out than the plan's full length, we delay the start rather than stretch phases thin. If it's closer, we compress — dropping optional phases first, then scaling — but never below the distance's pedagogical minimum. Below that, we decline to sell you a plan that can't do its job, and say so.

Your preview shows two numbers. Today's fitness is your current VDOT run through the Riegel model at race distance — what you could run if the race were this weekend. If your baseline result was at the same distance as your goal race, this number simply echoes it; that's the model being honest, not broken.

The headline is the trained projection: where a completed block typically lands you. We apply a conservative VDOT gain that shrinks as fitness rises —

gain/week = clamp(0.32 − 0.005 × VDOT, 0.03, 0.18)   (capped at 3 per block)

— roughly 1.2 VDOT points for a 24-minute 5K runner over a 10-week build (about half a minute at 5K), tapering toward zero at elite fitness, where blocks consolidate more than they add. The range spans a modest block (40% of the gain) to a strong one (the full gain). It's a model, not a contract: it assumes you run the plan, sleep, and arrive healthy. We'd still rather you beat the number than resent it.

The structure above — dates, distances, paces, phases — is pure deterministic math. The coach briefings on each workout (what to do, how it should feel, and why this session sits on this day) are generated from your plan's real numbers and automatically audited against one rule above all: never invent a pace, never contradict the plan. The math decides; the words explain.