In mathematics, a gradient is a vector pointing in the direction of the steepest descent/ascent. The magnitude of the vector is how steep the slope is, and so the gradient is the vector with the largest magnitude, or in other words the largest change. If we want to find the highest point in a topography (the maximum of the objective function, one approach is to always follow the direction of the steepest gradient, which will lead us to the local maximum faster than other methods.
The largest hurdle when working on a new project or learning a new skill is where to go first. Especially with skills, without context or structured education, we struggle to break down skills into smaller requisite sub-skills, and struggle to prioritize how much effort to allot to these sub-skills and in what order. Essentially, we cannot find the optima and learn the whole skill because we are unable to find the gradient of improvement. This can often be the difficulty in using a greedy algorithm.
One way to quickly find the gradient of improvement is to build a cognitive scaffolding from other developed skills.
This analogy is also useful when visualizing plateaus and local maxima. Just like in mathematical optimization, we require a more sophisticated method to get out of plateaus and local maxima which may involve temporarily reducing our level of skill to go in the right direction. Oftentimes this consists of working away bad habits.