FitSide

Most fitness programmes online are written for one person: a young adult with a full gym, no injury history and an hour a day to train. That person is rare. The people who actually search for a programme train at home, manage a chronic condition, are coming back from an injury, are considerably older, or cannot promise the same hours every week. The generic plan fails them at the first session it assumes they can do.

The alternative has always been a personal trainer, who would look at the whole person and write something that fits. That service is expensive and, for most people, out of reach. Between the free programme that does not fit and the trainer they cannot afford, there was nothing.

FitSide was built for exactly that gap. Its user knows what they want to achieve. What they lack is the professional who would tell them how to get there given their body, their schedule and their equipment. The product's job is to produce what that professional would produce.

The hard problem was making the personalisation real rather than cosmetic. It is easy to collect inputs and print them back as labels on a template. It is much harder to produce a programme that changes in substance when the inputs change: training experience, equipment to hand, minutes per session, sessions per week, the goal, the shoulder that cannot go overhead, a preference for how training should feel.

Those factors interact. Fewer sessions per week means each session carries more, which changes exercise selection and not just volume. An injury removes movements and forces substitutions that still serve the goal. A plan that gets one of those wrong is not slightly off; it is the wrong plan, and the user will know within a week.

Two further constraints sat on top. The programme had to be specific enough to follow without thinking and simple enough that nobody abandoned it. And it had to arrive in seconds, because a user who has just answered a set of personal questions does not wait around for a document to be assembled.

The intake is a structured flow rather than a form. It asks for the fitness goal, experience level, available equipment, sessions per week, time per session, and any injuries or areas to avoid. Each answer shapes the next, so a home trainer is never asked about cable machines. The finished set of inputs is the complete brief.

That brief goes to OpenAI, which returns a full weekly programme: exercises, sets, reps, rest periods and the progressions that carry the user from one week to the next. The output is structured as a written plan the user can follow session by session, not a block of prose to interpret. It appears within 90 seconds of submission.

Every exercise carries a description and form notes calibrated to the stated experience level. A beginner gets more detailed cues than someone who has been training for years, because the same instruction is either essential or noise depending on who reads it. If the first output does not suit, the user can regenerate, or adjust a specific input and run again. Each programme saves to the profile with the intake that produced it, so progression over time is visible rather than remembered.

The alternative was a large library of pre-built programmes behind a filter: choose a goal, choose your equipment, take the closest match. It is deterministic, reviewable by a coach in advance and faster to build.

FitSide's users are not alike, and the combinations multiply quickly. A 55-year-old training at home, two sessions a week, with a shoulder impingement and a goal of holding on to muscle mass does not map to any template a library could realistically hold. Filtering returns the nearest approximation and asks the user to live with the difference. Generation produces a programme for that exact combination.

The trade-off is control. A generated plan cannot be reviewed line by line before it exists, so the intake carries the load: the more precisely it captures constraints, the more precisely the output honours them.

A new user opens FitSide, works through the intake in around 3 minutes, and reads their first week's programme before the kettle has boiled. They train from the plan session by session, with the form notes beside each movement. When a week is done, the progression is already written into the next.

When something changes, they change the input. A new piece of equipment, a session lost to work, a knee that needs a few weeks off: adjust the field, regenerate, continue. There is no trainer to book, no reply to wait for and no template to search through. The profile keeps every programme and the intake that produced it, so the history of what was asked and what was prescribed sits in one place.

Nothing in the loop needs a human on the FitSide side. Assembly, cue calibration and history all happen without intervention, which is what lets a service shaped like personal training run without a trainer's calendar behind it.

FitSide launched in Melbourne with the full loop live. A user completes the intake in around 3 minutes and holds a full weekly training programme built for their inputs within 90 seconds of submitting it, well under 2 minutes from the last question to the first exercise.

People who found generic programmes too generic and personal training too expensive now have a structured, specific plan that accounts for their actual situation rather than an approximation of it. The foundation underneath is an intake that captures real constraints and a generator that honours them, which is what every later programme, regeneration and progression is built on.