Type 같이 into a romanizer and you may get 'gat-i'. Say 'gat-i' to a Korean and you'll get a polite pause while they reassemble what you meant, because every Korean alive pronounces that word gachi. Neither the romanizer nor your textbook lied, exactly. They transcribed the spelling. Korean, like English, doesn't pronounce its spelling — and the gap between the two is where most pronunciation frustration lives.
Spelling is etymology, sound is physics
Korean spelling preserves word roots so that meaning stays visible: 같이 is 같- (same) plus -이, and Hangul keeps that structure on the page. But when a mouth actually produces those syllables at speed, adjacent sounds collide and simplify — ㅌ meeting 이 palatalizes into a ch sound, and out comes gachi. The spelling shows you what the word is; the pronunciation is what a tongue can physically do at conversation speed.
English works the same way — 'don't you' comes out 'don-chu' — but nobody hands English learners a letter-by-letter transcription and calls it pronunciation. Korean learners get exactly that, constantly. Any romanization that goes syllable-block by syllable-block is transcribing the etymology, not the sound.
The five collisions that cover most of it
The good news: Korean sound changes are regular. A handful of rules cover the overwhelming majority of everyday words. Linking: a final consonant slides onto a following vowel — 맛있어요 flows as ma-si-sseo-yo, not mat-i-sseo-yo. Nasalization: stops soften into nasals before ㄴ and ㅁ — 한국말 becomes hangungmal, 감사합니다 ends in hamnida, not hap-ni-da. Palatalization: ㄷ/ㅌ plus 이 turns into j/ch — the gachi rule. Aspiration merging: ㅎ fuses with a neighboring stop — 어떡해 comes out eotteokae. Tensing: after certain stops, the next consonant hardens — 학교 sounds like hakkyo.
You don't need to memorize these as rules any more than English speakers memorize 'don-chu'. You need to know they exist, so that when audio contradicts romanization, you trust the audio. The rules then install themselves through listening, which is how Koreans acquired them too.
Why we still use romanization here — and how
Given all this, the purist position — no romanization ever — is tempting, and for reading Korean it's correct: Hangul takes an afternoon to learn and romanization read as spelling becomes a crutch that caps your pronunciation permanently. But romanization has one honest job left: telling you what a word sounds like before your Hangul reading is fast enough to hear it in your head.
That only works if the romanization transcribes the pronunciation, not the spelling. It's why every entry on this site romanizes the sound with the changes applied — gachi, eotteokae, hamnida — and why our romanization sometimes disagrees with the letter-by-letter version an automatic converter produces. When they disagree, ours is describing the word Koreans say; the converter is describing the word as written. Both are 'correct'. Only one helps you speak.
Treat romanization like training wheels with an expiry date. Read the Hangul first, glance at the roman only to check the sound, and listen to the audio last as the ground truth. Within a few weeks the middle step goes quiet on its own — you'll read 같이 and hear gachi directly, which is the moment romanization has done its job and can be ignored. The learners who struggle longest with pronunciation are almost never the ones who ignored romanization too early. They're the ones who trusted it too long.