Taylor Polynomials, Visualized: Rebuilding a Curve from One Point

If you only know a function at one point — its value, its slope, how it bends — how much of the curve can you rebuild? We build Taylor polynomials for eˣ at 0 one term at a time: the flat line, the tangent line, then a parabola with the same curvature. A derivative table shows that each new term xᵏ/k! fixes exactly one more derivative without disturbing the earlier ones, and zooming in shows why: halving the distance shrinks the error of T₁, T₂, T₃ about 4, 8 and 16 times. The general recipe follows: the coefficient is the k-th derivative divided by k!. At x = 0.5 the error falls from 0.649 to below 10⁻⁸ by order 8. Finally, the limit: for 1/(1 + x²), more terms help only between −1 and 1; outside, the polynomials swing ever wider, because the radius of convergence is 1. A finite Taylor polynomial is a local approximation, and convergence must be checked separately. Narrated in English with a synthetic voice. Every curve and number is computed with Python/NumPy and drawn with Manim. Chapters 0:00 How much can one point tell you? 0:19 WHAT — Start at one point 2:04 WHY — Each order fixes one behaviour 3:33 HOW — The general recipe 5:09 LIMITS — More terms are not always better 6:46 Summary and a practice problem Music: Cumulus Nimbus — Quincas Moreira (YouTube Audio Library)