Beta · Tartu, then everywhere

Every athlete has one thing holding them back.

A personal trainer writes your plan every two weeks. Terav sharpens the plan against your data — every rep, every note, every morning.

Free during beta. Beta users keep lifetime free access.
100+
primary studies
Every rep
adapts, not weekly
0€
during beta
Today · Tuesday
9:41
Terav
100
2 updates from yesterdaytap
Left hip stiffness noted 2× · load ×0.90 proposed
Engine Builder · Week 3 of 8
Aerobic base · Norwegian 4×4
HR 90-95% · row or ski-erg · fuelled
4 × 4 min interval
3 min active rest
Interval
Set
Time
HR
RPE
1
1:32
182
8
2
1:34
184
8
3
4
Note detected
“legs feeling flat, low fuel this morning”→ fatigue signal · Accept ×0.90?
TodayWeekProgressHistory
Where Terav sits

Not a log. Not a template. Not a monthly plan.

The honest three-way comparison. Every claim traces to a study or a design decision we’ll defend.

Strong · Hevy
Log what you did

They record your sets. That’s the whole job. Next Tuesday you have to figure out what to change yourself.

Terav
We change what you do next.
Fitbod · Freeletics
Pick a template

You get sessions from a library. Same session for a strong lifter and a returning novice with hip pain.

Terav
Every session is generated from your intake + last week’s log.
A personal trainer
Writes a plan every 2 weeks

Median trainers batch adjustments biweekly. Between sessions, you’re alone with what they wrote on Tuesday.

Terav
We adapt every rep, every note, every morning.
And Runna? Great if your one goal is a race. If your body has to last thirty more years past that race, Terav trains the whole system — and includes running programs when running is what you’re building.
How it actually works

From first question to fourth block.

Not a magic AI black box. A designed generator with a documented adaptive layer, showing its work at every step.

01

Intake

Nine questions plus a short physical assessment. What holds you back, what you’ve done, what hurts, what you have to train with. No template picker.

Traces to
Multi-dimensional capability profile (whitepaper Part 3). Per-capability level, not a tier bucket.
Intake · Step 4 of 9
9:41
Terav
100
Aerobic capacity
2 km row time trial — best in the last 6 months?
Best guess is fine. We’ll retest at week 4.
Under 7:30
High
7:30 – 8:30
Moderate
8:30 – 9:30
Building
Over 9:30 · haven’t rowed
Foundation
02

Personalised plan

The generator writes a plan against your profile and your history. Every session cites the study it’s built on. Every drill lists the external-focus cue verbatim.

Traces to
generateWeeklyPlan(program, profile, historicalLogs). Wulf 1998, 2013 external-focus cues on every drill card.
Week · Engine Builder block 1
9:41
Terav
100
Week 3 · Mon 11 → Sun 17
Aerobic block · concurrent strength
5 / wk
MonZone 2 · row 45 minAEROBIC
TueSquat + press · RPE ≤ 7STRENGTH
Lift first — same-day (Eddens 2018)
WedRest · walk 30 minRECOVER
ThuNorwegian 4×4 · ski-ergINTERVAL
Helgerud 2007 — 4×4 drives SV
FriDeadlift + row · RPE ≤ 7STRENGTH
SatZone 2 · row 60 minAEROBIC
SunRestOFF
Rowing over running for a strength athlete — less bidirectional damage (Doma 2019). Tap for source.
03

Every session adapts

You log a rep, an RPE, a one-line note. The engine reads it. It proposes a load shift, a substitution, or a rest day. You Accept or Ignore. Nothing changes without your Accept.

Traces to
Confirm-first — engine proposes, user accepts. Documented in note-signals.ts, suggest.ts, adapt.ts.
Today · Tuesday
9:41
Terav
100
2 updates from yesterdaytap
Left hip stiffness noted 2× · load ×0.90 proposed
Engine Builder · Week 3 of 8
Aerobic base · Norwegian 4×4
HR 90-95% · row or ski-erg · fuelled
4 × 4 min interval
3 min active rest
Interval
Set
Time
HR
RPE
1
1:32
182
8
2
1:34
184
8
3
4
Note detected
“legs feeling flat, low fuel this morning”→ fatigue signal · Accept ×0.90?
TodayWeekProgressHistory
04

Retest & continue

Every 4 weeks we retest the metric that matters — submax HR, hold time, the thing your program is actually chasing. Then we sharpen the next block against what we learned.

Traces to
Track threshold / submax HR / capability level as primary metric — not VO2max (Joyner & Coyle 2008).
Progress · Week 3
9:41
Terav
100
This week

Submax HR down 11 bpm at row-pace-5. Left hip flags dropped from 4 to 1. Load held. Read as adaptation, not fluke.

Hip flexor · left
trending
MonSun
Submax HR · row pace 5
−11 bpm
W1W2W3
The catalog

Programs, honest about what ships.

Two available now, four in build. Each one declares its evidence base and its engineering flags before you sign up.

Aerobic base · strength athletes
AVAILABLE

Engine Builder

8 weeks · block 1 of 3

For lifters who can barely run 5k. Row / ski-erg / bike Zone 2 plus a weekly Norwegian 4×4, with strength held at RPE ≤ 7. Expect −8 to −15 bpm submax HR by week 8.

Cites
Helgerud 2007; Wisløff 2007; Schumann 2022; Wilson 2012; Robineau 2016
Rehab-adjacent · pain-informed
AVAILABLE

Anterior Hip Rebuild

12-month arc

The programme Margus built for himself after a labral question mark and years of anterior hip pain. Symptom-gated progressions, no forbidden movements at the wrong phase.

Cites
Sands 2000; Gabbett 2016; supplement to physiotherapy — not a replacement
Skill · multi-dimensional
COMING

Handstand Walk

Multi-tier · beginner → advanced

First-hold through obstacles. Six independent capability tracks (hold, walk, turns, wrist load, shoulder endurance, obstacle) each sharpened at its own level.

Cites
Wulf 1998, 2013; Shea & Morgan 1979; Karni 1998; Sleeper 2012 GFMT
Strength · component build
COMING

First strict pull-up

8-12 weeks

Scap control, dead-hang tolerance, negatives, band-assisted, then unassisted. Each sub-skill tracked and progressed independently.

Cites
Motor learning primary sources · specificity of practice
Skill · wall to freestand
COMING

First strict HSPU

12-16 weeks

Wrist load tolerance first, then wall-facing hold, box HSPUs, kick-up work. Same multi-dimensional generator as the handstand walk arc.

Cites
Wulf & Shea 2002; Robertson 2004 consolidation windows
Race prep · concurrent
COMING

HYROX prep

12 weeks · race-anchored

For the roxer who wants to keep squatting. Modality selection to protect the strength floor; VO2max is the strongest HYROX predictor.

Cites
Brandt 2025 HYROX profiling; Schumann 2022; Wilson 2012
The evidence claim

100+ primary studies. Every session cites its research.

This is the trust lever. Every rule we’ve baked into the engine — RPE ceilings, session order, modality, feedback type — traces to a peer-reviewed source. When we don’t have one, we label the choice as engineering, not evidence.

Aerobic
25+

Mitochondrial biogenesis, capillary density, stroke volume, HR zones, individual variation.

Concurrent
35+

Interference effect, session order, modality, protein dose, hybrid athlete research.

Motor learning
40+

External focus, contextual interference, feedback type, consolidation, deliberate practice.

Read the citations →Grouped by domain. Effect sizes where the literature has them.
Specialist report · Nov 2026
9:41
Terav
100
Report · 6 weeks · Terav
Left hip / anterior · training-load cross-ref
Symptom log
Oct 04click on lowering, 3/10 · leg raises
Oct 09ache post-run, 2/10 · settled overnight
Oct 15no flag
Oct 22mild stiffness AM · gone by warmup
Nov 05no flag · 2×squat session · loaded
Load context
Sessions
14 / 18
Skipped
0
Avg RPE
6.8
Adj. load
×0.94
Interpretation
Anterior click on lowering phase of hanging leg raises resolved by Oct 22 after removing eccentric hip flexion (see substitution log). No return in following 2 weeks at higher squat load.
Honest section

What we won’t do.

An explicit refusal list. Because the app makes strong claims where it can, the honest limits belong right here in the front door.

  • We won’t fake live form check

    Your form isn’t visible to a phone. Drill cards flag common failure modes and force honest self-report. The notes engine catches the drift a screen can’t see.

  • We won’t fake medical assessment

    The app is a training log. Red-flag patterns fire an escalate-to-clinician banner — not a diagnosis, not a treatment plan.

  • We won’t fake certainty about you

    VO2max response has a ~10× range across individuals in the peer-reviewed record. Programs quote expected outcomes with the honest ranges from the literature, not one number.

  • We won’t fake real-time cueing

    A trainer’s live “elbow higher” doesn’t come through a screen. What we ship is Wulf’s external-focus cue on every drill card — the evidence-based verbal substitute.

  • We won’t fake nutrition coaching

    We surface the peer-reviewed protein / carb reference (Morton 2018, Impey 2018). We don’t prescribe meals or track your macros. That’s a different app.

  • We won’t guilt you to open the app

    No dark-pattern streak notifications. If you skip a week, the plan sharpens against that too. The specialist report is the proof the arc is real — not a red badge.

Terav · Estonian for sharp
“A blade gets sharper by grinding against something harder than itself. So does a training plan.”

Terav is Estonian for sharp. The plan gets sharper each session against the hardest thing available — your data. What you did yesterday, what you noted this morning, how your last block actually went.

Built by athletes at a real CrossFit box in Tartu, out of a multi-year rehab log for one lifter with a hip that wouldn’t quit acting up. If we can sharpen a plan against that, we can sharpen one against yours.

Sharpen the plan.
Every session.

Free during beta. Backed by 100+ studies. Built by athletes at a real CrossFit box in Estonia. Beta users keep lifetime free access.