A few years ago, a tech company released a product called Grok. The internet immediately divided into two camps: people who pronounced it grock, and people who pronounced it groak. The company clarified that it was grock. They had to clarify. Because the letters they chose to represent the word did not, on their own, succeed in representing the word.
Ask yourself what kind of writing system requires the inventor of a word to separately announce how it is pronounced. That is not a writing system. That is a memory system with a typing interface.
A Problem Italian Cannot Have
Imagine an Italian startup launches a product called Grok. Every Italian speaker reads it the same way: grok. There is no alternative. The spelling-to-sound mapping in Italian is deterministic. You see the letters. You know the sound. End of process.
The same is true in Finnish. In Spanish. In Turkish. In essentially every language that uses an alphabet for what an alphabet is for—encoding sound in a predictable way. If you invent a new word in any of these languages and write it down, the reader will pronounce it correctly without ever having heard it. That is what an alphabet is. That is the contract.
English does not honor this contract. English has been violating it for so long that we have stopped noticing. And the startup industry has weaponized the violation.
The New-Name Hall of Shame
Consider a small sample of company and product names from the last decade:
| Name | What the letters suggest | How you are actually supposed to say it | The trick |
|---|---|---|---|
| Grok | groak? grock? grohk? | grock | Short O like rock, but the word looks like broke |
| Claude | klawd? klaw-dee? | clawed | Final e does nothing, au is French, you were supposed to know that |
| Lyft | lift? lieft? | lift | y standing in for i, contributes nothing but a logo |
| Flickr | flick-er? f-like-er? | flicker | Missing e—you fill it in from context, hope you are right |
| X (the platform) | eks? zhee? ha? | depends who you ask | One letter. Zero phonetic information. Maximum branding. |
Each of these names exploits a different failure mode of English orthography. Grok exploits vowel ambiguity. Claude exploits silent letters and foreign borrowings. Lyft exploits the letter y doing whatever it wants. Flickr exploits the fact that English speakers are trained to mentally insert missing letters because the language has conditioned them to expect nothing from its spelling. X exploits the fact that at this point, nobody expects a letter to correspond to a sound anyway.
The companies are not breaking the language. They are using it exactly as designed. The design is the problem.
A Language That Cannot Be Violated
There is a useful concept in information theory: error-correcting codes. A good code is one where small errors do not produce a different valid message. If I type thare instead of there in English, I have produced a misspelling—but the word is still decipherable. The system has some built-in redundancy.
Now apply this to pronunciation. In a well-designed orthography, a slight deviation in spelling produces a slightly different but still valid pronunciation—or, better yet, the same pronunciation. In Italian, if you misremember whether a word is spelled with one t or two, it barely matters, because double consonants are a systematic marker. The rule tells you what to do.
In English, the same situation produces chaos. Through and though are one letter apart. One rhymes with blue. One rhymes with go. There is no error correction. There is only right and wrong.
This makes brand names in English a unique form of social friction. Every new name is a gamble: will people guess the pronunciation, or will the company have to issue a press release explaining what its own name sounds like? The fact that companies routinely do the latter is not a cute quirk of startup culture. It is a diagnostic signal that the underlying writing system has failed.
It Does Not Have to Be This Way
No one is asking startups to name their products boring things. You can call your AI Grok. You can call it Claude. But in a language with a functioning orthography, the name would carry its pronunciation with it, and no press release would be necessary. The name would work the first time.
A writing system that cannot reliably encode a five-letter made-up word is a writing system that has stopped doing its job. English stopped doing its job a long time ago. We just keep hiring it back out of habit.
When an Italian child sees a new word, they read it. When an English-speaking adult sees a new product name, they guess, then Google it, then watch a YouTube video where the CEO says it, then still get it wrong at the next meeting. This is not a personal failure. It is a system failure, committed by the language, paid for by the user, and monetized by branding consultants who have convinced companies that unreadable names are cool.
They are not cool. They are evidence, written fresh every quarter, that your writing system has failed.