Generate a physique projection

Three months into training, you're trying to imagine where this is heading — twelve months out, what's the realistic body if the work holds. You ask the agent to project from your current data — pessimistic, realistic, and optimistic scenarios, each grounded in what's actually been happening.

What the agent needs to hear

Behind your sentence, the agent works backwards to two pieces: the horizon and what scenarios to render. The horizon defaults to 12 months — far enough to matter for planning, close enough that current trends still carry signal. Override with any duration you’d say out loud: « 6 months out », « project two years », « where am I by next summer ».

Scenarios default to three side-by-side — pessimistic, realistic, optimistic. The agent reads your training history, body data, nutrition adherence, and progression rate, then renders each scenario with grounded assumptions instead of made-up numbers. Silent on both, you get 12 months across all three.

What goes into the projection

Four inputs shape each scenario. Training progression rate reads how fast strength has been moving — the realistic curve extends that pace, the optimistic assumes it holds, the pessimistic assumes it halves. Body trend picks up weight and perimeter trajectories from the logs you’ve kept. Nutrition adherence reads on-plan vs off-plan when a diet plan has been active. Age and starting point anchor the realistic curve — a beginner’s first year is not a tenth-year lifter’s plateau, and the projection knows the difference.

The pessimistic scenario assumes minimal progression and patchy consistency. The realistic extends current trends without heroics. The optimistic assumes peak compliance and the upper end of typical progression for your profile. Three reads, one set of inputs.

The projection is a thought tool, not a forecast

A Bayesian best guess across three confidence bands. The projection is a structured imagination exercise, not a prediction. Real life rarely lands on any single scenario; the value is seeing the spread.

The pessimistic answers « what if I half-ass this? ». The realistic answers « if the work continues at this pace, where does it go? ». The optimistic answers « what’s the upper bound if I show up every week? ». Use the set for goal-setting and honest framing — not for accountability or self-judgment. The projection gets saved so you can revisit it in six months and see which scenario you actually tracked.

When the agent gets it wrong

Three patterns show up. Scenarios feel implausible because the data was thin — three months of sparse logs can’t carry a twelve-month projection cleanly. Narrow the scope: « project only the lifts I’ve trained consistently ».

The realistic curve assumes a progression rate that doesn’t match recent slowdowns. The agent averaged across the full window when the last 8 weeks tell a different story. Re-anchor: « weight the last 8 weeks heavier than the first 8 ».

The horizon is too long for the data underneath it. A 12-month projection off 3 months of training is a guess pyramid — the further out, the wider the error bars get, until the optimistic and pessimistic stop overlapping with reality. Trim: « narrow to 6 months, the data isn’t deep enough for a year ».

What makes the projection worth pulling

Three things decide whether the projection helps you plan: the horizon matches the planning window you actually care about (a 5-year projection is for thinking, a 12-month is for goal-setting, a 6-month is for accountability), the scenarios are read as ranges, not predictions (the realistic scenario isn’t « what will happen » — it’s « what would happen if current trends extend »), and the projection gets revisited (saving it and looking at it six months later, comparing to what actually happened, is what makes the exercise useful). Saved projections become reference points for honest mid-block check-ins.