Safe by design
The AI can always go lighter. It can only go slightly heavier. Anything outside the safe range is rejected, and the engine’s number stands.
Under the hood
How our AI is built.
Two minds and a referee. A rules engine built on sports-science research, an AI coach that reads your whole story, and tested code that checks every decision before you see it.
01 · What it reads
- Your sets, reps and how each one felt
- Today’s check-in: sleep, soreness, time
- Recovery estimate for every muscle
- Your history, and what you’ve told it
02 · The brain
Rules engine
Built on published research. Runs on your phone, even offline. Works out your next set and a safe range around it.
AI coach Pro
Reads your whole story and suggests a better next set, inside that safe range.
The referee: tested code checks every suggestion. Outside the range? The engine’s number stands.
03 · What you see
NEXT · SET 3
57.5 kg × 10
Coach adjusted
Why? Slept 4 hours and that one was grindy. Stay at 57.5, aim for 10.
Your hands always win: touch the weight and the suggestion steps aside.
Never waits for the network
The engine answers instantly on your phone, even with no signal. The AI’s suggestion arrives while you rest, or not at all.
Remembers you, in your words
It remembers what you tell it, like “night shifts” or “left shoulder tight since October”. It doesn’t guess about you, and forgets when you ask.
Safety check first
Words like chest pain or dizziness are caught by a safety check before the AI even answers. Every coach gives the same calm advice: stop, and see a doctor.
- 5
- hard limits checked in code on every plan
- 6
- coach personalities, one shared brain
- 30
- test scenarios run on every coach before an update ships
- 0
- bars of signal needed for your next set
Lifetime and offline: the rules engine runs every decision on its own. Pro adds the AI coach on top, always inside the same limits.
The AI model
Powered by Claude, from Anthropic.
We’re proud to build our AI coach on Claude, made by Anthropic, one of the world’s leading AI labs. Anthropic puts safety at the centre of how it builds AI. That’s the same way we built hustl.fit: the AI gets your whole story, and our tested code keeps every decision inside safe limits.
- Before launch, we test Claude models against our 30 coaching and safety scenarios and pick the one that passes them all.
- Your health data (weight, injuries, meals) never goes to analytics or ads.
- No AI? No problem. Offline or on Lifetime, the rules engine makes every decision on its own.
Claude is a trademark of Anthropic, PBC. hustl.fit is an independent app and isn’t affiliated with or endorsed by Anthropic.
The science
The research behind the rules.
Every rule in the engine comes from published sports-science research. Where a number is our own calibration on top of the research, like the recovery formula, the app says so and calls it an estimate.
- Progressive overload and weight steps
- ACSM position stand, Progression Models in Resistance Training for Healthy Adults (2009)
- How a set felt → how hard it was
- Reps-in-reserve effort scale: Zourdos et al. (2016); Helms et al. (2016)
- Rep ranges, weekly sets, training each muscle twice a week
- Schoenfeld and colleagues (2016, 2017, 2021); ACSM (2009)
- Rest between sets
- Schoenfeld et al. (2016); ACSM (2009)
- About 48–72 h before training a muscle again
- ACSM (2009); muscle protein synthesis studies (MacDougall et al. 1995; Phillips et al. 1997)
- Daily calories
- Mifflin–St Jeor equation (1990)
- Protein, 1.6–2.2 g per kg
- Morton et al. (2018) meta-analysis; ISSN position stand on protein (Jäger et al. 2017)
- Indian food values
- Indian Food Composition Tables 2017, ICMR-NIN Hyderabad