A set has measure zero (is null) if μ(A) = 0. A property holds almost everywhere (a.e.) if the set where it fails is null. This allows us to ignore 'small' sets and treat functions differing on a null set as equivalent.
Observe that single points in ℝ have Lebesgue measure zero, as do all countable sets. See how L^p spaces identify functions equal almost everywhere.
Null sets are not necessarily empty; the Cantor set has measure zero but is uncountable. 'Almost every' quantification must be formalized via measures, not classical logic.
From your prerequisite, you know that a measure space is a triple (X, Σ, μ) where Σ is a σ-algebra of measurable sets and μ assigns non-negative sizes to those sets. A null set is simply a measurable set A ∈ Σ with μ(A) = 0. The name suggests these sets are "negligible" or "invisible" to the measure, and that intuition is exactly right: a null set contributes nothing to any integral and can be safely ignored in most analytic arguments.
The key example to anchor this is Lebesgue measure on ℝ. A single point {x₀} has Lebesgue measure zero, as does any finite collection of points, and by countable additivity, any countable set — including all the rational numbers ℚ, which are dense in ℝ but form a measure-zero set. This already shows that "measure zero" does not mean "sparse" in the topological sense: ℚ is dense yet negligible. More dramatically, the Cantor set is an uncountable subset of [0,1] with Lebesgue measure zero. Null sets can be large from a cardinality perspective while remaining invisible to integration.
The phrase almost everywhere (a.e.) means "at every point except possibly a null set." If a property P(x) holds a.e., then {x : P(x) fails} is a null set. For example, we say two functions f and g are equal almost everywhere (f = g a.e.) if {x : f(x) ≠ g(x)} has measure zero. This is the key equivalence relation underlying Lᵖ spaces: in L²([0, 1]), the functions f(x) = 0 and g(x) = 𝟏_{x=1/2}(x) (which differs from zero only at one point) are identified because they differ only on a null set.
The practical power of the almost-everywhere concept is that it allows you to ignore countably many exceptional points — discontinuities, singularities, individual bad values — without affecting integrals or limiting arguments. When you prove that a sequence of functions converges, saying it converges a.e. is often the strongest natural statement you can make. The big convergence theorems you will see next — such as the Dominated Convergence Theorem — operate in this language: hypotheses and conclusions are stated a.e., which is exactly the right granularity for Lebesgue integration.