Mergers of formerly distinct sounds that are spelled differently are easier to notice than splits of sounds. There are examples of splits that are introducing new contrasts into English phonology.
One phoneme prone to splitting in a few regional accents is the front low vowel /æ/. It is usually analyzed phonologically as a "short" or "lax" low vowel. But for some speakers, there are a few minimal pairs or near-minimal pairs between words with short/lax [æ] and words with a lengthened variant [æː] or "tense" variant [eə]. For more information, look up "bad-lad split".
At the level of whole languages, "phonological complexity" is not a simple concept because phonologies are not directly observable. One analysis might have half as many phonemes as another, depending on whether certain sounds are analyzed as sequences or phonological units. New phonological units are often historically derived from older sequences, and the transition from a sequence to a phoneme can be hard to identify. E.g. in French, "nasal vowel" phonemes historically derive in most cases from a sequence of vowel + nasal consonant, and in certain respects they can be argued to behave like such sequences even in the phonology of modern French. In American English, "rhotic vowels" as in car, chair, core, cheer are conventionally analyzed as sequences ending in the consonant /r/, but they could instead be analyzed as unitary vowel phonemes, like diphthongs.
So in many cases, the development of a new phonological distinction is not clear until centuries have passed. Some examples of distinctions in modern English that we know developed from sounds that were not distinct in an earlier language/stage of the language: /k/ vs. /tʃ/, /f/ vs. /v/, /θ/ vs. /ð/, /s/ vs. /z/, the three-way split between /ʌ/, /ʊ/, and /uː/.