Strength training designed for women's bodies, one session at a time, for decades.

Women are told to choose. Train for longevity and you get a cautious, general program. Train for shape and you get a program written for men with lighter weights. EVLVE's position is that for a healthy woman in her late twenties to mid forties, the two are mostly the same work done carefully: heavy hip and leg loading protects bone and builds the glutes; a strong back and shoulders carry a long spine and widen the frame the waist sits inside; balance and impact practice are the part of longevity nothing else trains.

So every session carries both. A session is never "the bone day" or "the glute day". It is one hour that leaves nothing out.

Version 2.0 began with two audits of the live program, not a blank page. The first read every piece of programming logic in the database against the current evidence on women's strength, bone density, balance, mobility and hypertrophy: the rep ranges, the rest timer, the warm-up and cool-down, the order of a session, the progression rules, the substitutions. The second repeated the exercise for every equipment option a member can select, all 2,049 combinations the old plan could produce, and found where a home member with two dumbbells and a band was being given something the evidence does not support.

Both audits were fact-checked by a separate pass that had not written them. Numbers that did not survive were corrected in the report, not softened.

Most apps have a few templates and a swap button. EVLVE resolves every slot of every session from the member in front of it: what she owns, which weights and bands exactly, where she trains, her level, her time, where she is in the program, and what her body needs worked around. The space that produces is large, and it was sized on purpose.

The audit found what most apps hide: a member who logs exactly what she is shown could be walked down to a fraction of her starting weight over a few months, by rules that read a missed set as a reason to retreat. The rebuilt engine was written so that cannot happen, and 287 test vectors prove each rule, each one paired with the defect it removes.

• Real laddersThe next weight is the next weight she owns: her dumbbell pairs, her plates on her bar, her stack's real step, her band by kind. Nothing is prescribed that is not in her list.
• A jump is predicted before it is triedWhere her pairs jump 15 to 20 lb, the engine predicts her reps at the heavier weight from her best recent set and only offers it when the prediction supports five. Until then she bridges: more reps, a slower lowering, then the harder version.
• Never a max set on a barThe one calibration set that asks for as many good reps as she can is allowed on dumbbells, bands, machines and bodyweight, and never on a barbell or trap bar.
• A step down is immediate and honestTwo misses cancel a step-in and return her to the lighter weight at her old target, with nothing recorded as a back-off. Lighter rotations count toward no stall clock.
• Back after a breakA per-lift re-entry table by days away, using the conservative end of each range, offered as "lighter today, on purpose". The app never says how long she was gone.
• Bodyweight has a ladder tooNinety-seven variation families and 261 rungs: a push-up member moves from incline to floor to deficit; a pull-up member from hangs to band-assisted to the bar, with a chair and a step as the honest requirements.

EVLVE asks nine plain questions about what to work around: knees, lower back, wrists, shoulders, hips, neck, pelvic floor, grip, and whether she wants a low-impact track. Each one maps to a rule set over every exercise's position tags, 66 rules in all: avoid, modify or prefer, with the cue that goes with it. A member with a wrist consideration gets dumbbells held neutral and a push-up on handles, not a shorter session.

The pain rule is a rule, not a disclaimer. Pain on an impact rung drops her to the low-impact track for the session, holds any step-up for two rotations, and shows the one signpost that applies. Leaking on impact does the same and points to pelvic-floor physiotherapy. Nothing is named back to her, nothing is diagnosed, and the second tap that asks which it was is the only question added.

What she is not asked: her cycle, her weight, her goal weight, or how she feels about her body. The evidence does not support programming by menstrual phase, and the app does not frame her as a project.

• Lights downOne signature move. She taps Begin and the porcelain page goes black over 900 milliseconds; the session title stays where it was and turns porcelain. Nothing slides. The light changes.
• One exercise, one screen, nothing movesA label, the name, her number at 72 points, one cue, one button whose label says exactly what will be logged. The bar and the button never change place between steps.
• Apple text styles, all the way downEvery size in the app is one of Apple's text styles and follows her Dynamic Type setting. The display number follows Large Title's curve; at accessibility size 3 the 72 becomes 110 and still fits a 375-point column with its unit wrapped beneath. Layouts reflow and never truncate.
• Three tabs, no iconsToday, Record and You, as words. No icon library, no emoji; the few marks the app needs are drawn as 1.4-point strokes.
• Nothing roundedSquare check boxes, square toggles, square sheets with a title and a close. The only rounded shapes near EVLVE are the system's own containers.
• Skipping is one tap and loses nothingNo sheet, no choices. Before the cool-down she sees what is waiting, each with "Go back to it". At the finish a skipped lift "stays". The words missed and incomplete are not used.
• Adequacy, not deficitNo scores, streaks, rankings, percentages or weight-loss projections anywhere. The record shows real numbers she did. A personal best is written on bordeaux, once, and bordeaux is used for nothing else.
• Reduce Motion, honouredEvery move becomes a 200-millisecond cross-fade or nothing. Nothing loops, nothing slides in on scroll.

Chandler Grant is 29 and had never shipped software before EVLVE. She taught herself the stack, wrote the product decisions, directed the research and the design pass, and did it alone, from St. Petersburg, Florida. The research and the engine specification were produced with Claude; the design pass was done in Claude Design; the code was written with AI coding tools against a design canvas that is the single source of truth. Every decision in this document is hers.

What one person with those tools could do in a year is part of the story: a research programme a sports-science department would recognise, a progression engine with a full test suite, and an interface held to a stricter standard than most teams manage. It is also a story about care. The first thing cut from every screen was anything that could make a woman feel managed.

• PlatformiPhone first. Capacitor over a React app, with the session player, haptics and the status bar treated as native concerns.
• ReleasesWeb changes ship as live updates with rollback; anything touching purchases or native capability goes through App Store review. 2.0's program rebuild cuts members over per engine version, with per-member rollback.
• DataRow-level security and explicit grants on every table a member touches, checked by a migration-time assertion. Analytics carry counts and durations only, never a consideration, a gap or an impact-check answer.
• Her first sessionFree, shorter by default, and complete. Onboarding is anonymous and asks the minimum; the detail comes after the plan is shown.