AN INTERACTIVE EXPLORATION

See the reason.

An equation is more than a sequence of steps. Make it move. Make it click.

01
THE INTUITION SERIESAlgebra & geometry
Completing the squareLIVE EXPLORATION
ƒ(x) =
Use x, fractions, + − * / and ^. Try
Your function
↔ Drag the vertex. Watch the algebra follow.
Vertex (2, −1)

FOLLOW THE WHY

A different form.
The same function.

≡
✓ Every step checked exactly
↳

THE AHA MOMENT

The vertex was hiding in the equation.

A LITTLE CHECK-IN 01 / 03

Move the vertex one unit to the right.
Which part changes?

Make a prediction. Then test your intuition in the graph.

THE FORMA FIELD GUIDE

Play with the idea.
Trust the reasoning.

Explore functions, equations, matrices, probability distributions, and sample data. Every symbolic claim names its rule and its conditions. Graphs and point values are approximate; they are never the proof.

  1. Topology studio accompanies selected sections of Munkres, second edition, with original explanations. Finite spaces on up to five points receive exact open-set, closure, continuity and separation checks; drawings do not prove general theorems.
  2. Power & planning uses known-sigma normal models for one mean or two independent means with equal group sizes. Numerical power and minimal sample sizes depend on the declared effect and assumptions.
  3. Bootstrap intervals resamples means, medians and paired mean changes with a saved seed. Percentile and basic intervals are approximate; resampling more often does not repair an unrepresentative sample.
  4. Function atlas searches 38 starting points across algebraic, rational, growth, trigonometric, inverse, hyperbolic, piecewise and composed functions. Graphing, differentiation and integration have separate capability labels.
  5. Probability lab explores binomial, Poisson, normal, uniform, and exponential distributions. Move event bounds and change parameters. Binomial masses are checked with exact fractions. Other distribution formulas name their assumptions and sources; numerical probability estimates are labeled separately from exact checks. A density is not the probability at a point.
  6. Bayes lab builds conditional probabilities from a joint count table or a prior and likelihoods. Select a conditioning event in the mosaic. Exact partitions and Bayes identities are checked; conditioning on zero mass is undefined.
  7. Statistical inference connects mean t/z procedures and proportion Wilson intervals/binomial tests to their assumptions. Change confidence, the null value and test direction. The displayed confidence interval stays two-sided. Numerical critical values and p-values are labeled estimates; p-values are not probabilities that a hypothesis is true.
  8. Compare samples supports Welch and pooled independent-mean tests and paired after-minus-before changes. Move paired values and the null difference; follow exact variance substitutions and numerical Student tests. Pooled tests require equal population variances.
  9. Regression inference adds slope/intercept tests, pointwise mean-response confidence intervals and wider individual prediction intervals. Move observations or the prediction x. Model assumptions stay explicit; perfect fits and unidentifiable slopes have no inferential certainty claim.
  10. ANOVA lab separates within-group and between-group variation using exact sums of squares. Its numerical F test asks whether all group means are equal under independent normal errors and a common variance; it does not identify which pairs differ.
  11. Post-hoc comparisons retains every pair of groups for pooled/Welch mean tests or Dunn rank comparisons. Exact contrast and rank steps accompany numerical tails. Holm is the default correction; the full family stays visible.
  12. Multiple testing compares raw and adjusted p-values with Bonferroni, Holm, BH or BY. Family-wise error and false discovery rate have different targets and assumptions. Effect-size panels in the comparison and omnibus labs describe magnitude using explicit conventions; they are not confidence intervals.
  13. Chi-square lab compares observed and expected counts for a fixed goodness-of-fit model or a contingency-table independence model. Exact expected-count and statistic identities are separate from the approximate chi-square tail. Sparse expected counts carry a warning.
  14. Rank tests supports Mann–Whitney, paired Wilcoxon signed-rank and Kruskal–Wallis. Ties receive exact average ranks. The first two use exact conditional permutations or sign flips within bounded datasets; Kruskal–Wallis uses a tie-corrected statistic and approximate chi-square reference. Nonparametric does not mean assumption-free.
  15. Statistics lab connects draggable observations to exact means, medians, spread, and quartiles. Paired data reveals least-squares regression and residuals. Population and sample variances have different denominators. Correlation and square roots are numerical displays; a fitted line does not establish causation or an inferential conclusion.
  16. Differential equations solves supported first-order initial-value problems: linear equations with polynomial forcing and autonomous quadratic growth. Drag the initial point over a slope field. Every solution carries its checked initial condition and connected solution interval.
  17. Linear algebra transforms vectors and a unit circle using exact rational 2 × 2 matrices. Explore determinants, inverses, eigenvalues, and row operations with exact checks behind the approximate picture.
  18. Complex plane colors output phase and magnitude. Principal branch values and analytic derivative conditions are distinct. Conjugate and modulus maps remain numerical explorations without an analytic derivative claim.
  19. Multivariable lab plots approximate level sets with a gradient arrow. Exact partial derivatives and Hessians retain source, gradient, and second-derivative domains.
  20. Special functions explores gamma, loggamma, digamma, erf, erfc, besselj0 and besselj1 with rational affine arguments. Linked theorem instances justify the supported derivative rules; bounded numerical methods supply curve values.
  21. Quadratic studio connects the vertex to exact completing-the-square identities.
  22. Trigonometry explores sin, cos and tan in radians. Follow identities and see where tangent is undefined.
  23. Logarithm lab expands supported log expressions with explicit positivity and nonzero conditions. Here log means natural logarithm, the same as ln.
  24. Integral studio finds supported elementary antiderivatives and checks them by symbolic differentiation. Drag the bounds to see signed area. A definite value is shown only when continuity is certified throughout the interval.
  25. Derivative lab applies sum, product, quotient and chain rules, including roots, real powers, inverse, hyperbolic, logarithmic and exponential functions. Boundary and corner exclusions remain visible.
  26. Identity checker uses exact algebra and supported identities. “Not verified” means the rule engine could not establish equality; it does not mean the expressions differ.
A bounded, explicit rule engine

Use x, rational numbers, pi, e and arithmetic, plus sqrt, cbrt, sin, cos, tan, sec, csc, cot, asin, acos, atan, sinh, cosh, tanh, asinh, acosh, atanh, ln, log2, log10, exp, abs, floor, ceil and sign. Write function arguments in parentheses. Integer powers accept negative bases where defined; noninteger and variable powers use positive bases. Use cbrt(x) for real cube roots of negative inputs. Plot-only functions do not receive a derivative certificate. The complex, multivariable, differential-equation, matrix, probability and statistics labs use their own documented grammars. Probability uses named distributions with bounded rational parameters and inclusive event bounds from −100 to 100. Statistics accepts up to 50 rational observations or x:y pairs, with each value between −10000 and 10000. Quartiles use medians of the lower and upper halves, excluding the overall median for an odd sample; the single observation supplies all quartiles for n = 1. Bayes accepts table(ab,aNotB,notAB,neither) or bayes(prior,sensitivity,falsePositive). Inference accepts mean(mean,s,n), zmean(mean,sigma,n), or proportion(k,n), with sample sizes up to 10000. Its confidence intervals are two-sided; selected test direction applies only to the hypothesis test. Compare samples accepts welch(m1,s1,n1,m2,s2,n2), pooled(m1,s1,n1,m2,s2,n2), or paired(before:after,...). Independent counts range from 2 to 10000 per group; paired and regression datasets contain at most 50 pairs. Regression inference requires identifiable coefficients, at least three pairs and positive residual variation. Response bands are pointwise, not simultaneous. ANOVA accepts anova(group values;group values;...) with 2–8 groups and at most 50 observations. Chi-square accepts gof(observed counts;positive weights) or independence(row counts;row counts;...). Rank tests accept mannwhitney(group1;group2), signedrank(before:after,...) and kruskal(groups separated by semicolons). The test-level control sets alpha; these three labs do not construct confidence intervals. Two-proportion tests, Bayesian credible intervals and arbitrary probability models remain outside this version. The differential-equation solver supports only its listed first-order families; higher-order equations, systems and numerical-only solutions are not included. Matrix calculations currently use 2 × 2 real rational matrices. Special-function integration, complex integration, multivariable integration, arbitrary piecewise syntax, improper integrals, and arbitrary proofs remain outside this version’s scope. Numerical ranges are distinct from mathematical domains. Domains survive cancellation. An antiderivative’s + C is independent on each connected domain interval.

Experiments save in this browser when storage is available. Share links include expressions and sample data, both point coordinates, integral bounds, and probability event bounds. Conditioning events, confidence levels, null values, test alternatives, regression targets, prediction x values, and field/grid visibility are saved too. No telemetry or AI calls. Exports are inspectable, unsigned verification records.