The science
Five ideas behind the coaching, and how the app and the AI use them. Each one is graded for how well the research supports it.
Your own data beats formulas
Consensus
Standard equations for energy use miss by more than 10% for a lot of people. PixelPT starts from a formula, then works out your real maintenance from what you eat and how your weight moves.
Frankenfield et al., 2005.
The trend matters, not the scale
Consensus
Day-to-day weight is mostly water. PixelPT smooths it into a trend, and marks expected swings such as salt, hard sessions and your cycle, so they don't trigger a calorie cut.
Walker, 1991. Kanellakis et al., 2023.
Protein first, and slow cuts
Consensus
Protein is the anchor of every target. Losing about 0.5 to 1% of body weight a week keeps more muscle than going faster, so that's the pace PixelPT guides you to.
Morton et al., 2018. Garthe et al., 2011.
A floor on how low it goes
Consensus
Eating too little for too long harms hormones and bone. Calculated targets keep an energy-availability floor, and pace advice tightens as you get leaner.
Nattiv et al., 2007. Mountjoy et al., IOC.
Honest about the weak spots
Contested or expert opinion
Carb timing, reverse dieting and our own lifestyle-context explanations are less proven. PixelPT labels them as such. It doesn't claim carb cycling melts fat or that reverse dieting repairs metabolism.
Impey et al., 2018. Trexler et al., 2014.
Who does what.
Three parts, each doing the job it's good at.
The research sets the rules
Targets, paces and floors come from published work, not from guesswork or a language model's habits.
The code does the maths
Trend weight, expenditure, calorie and macro targets, loads and goal progress are calculated by code and handed to the AI.
The AI does the coaching
It picks exercises, explains why, and writes in plain language. It never works out a number itself.