The idea
Software represents precise steps for transforming inputs into outputs. Good software also handles uncertainty, failures and the person using it.
An algorithm is like a recipe; a robust program knows what to do when ingredients are missing.
How it works
Algorithms
Procedures should be evaluated for correctness, time, memory use and behaviour on unusual inputs.
Symbolic computation
Symbolic mathematics manipulates expressions, whereas numerical methods approximate answers with finite precision.
Interfaces
An accessible interface makes what a system does understandable, reversible and testable.
Under the hood
Numerical computations require an error model, defined input ranges and precision. Testing should cover both expected outputs and failure conditions.
One small experiment
Compare the exact derivative of x² with a slope estimated from two nearby points.