Spellcheck is one of the most successful pieces of software ever built. It solves a clearly defined problem, it solves it reliably, and it has been built into almost every text input system in use today. If you type "recieve" instead of "receive," spellcheck catches it. Billions of typos and misspellings corrected, quietly, without friction.
The problem is that the errors spellcheck catches are almost never the errors that matter most for non-native writers in professional contexts. Spellcheck was designed for native speakers making mechanical typing errors. The patterns that affect non-native writing are different in kind, not just in degree.
What spellcheck was built to do
Spellcheck works by comparing each word in the text against a dictionary of correctly spelled words. When a word does not match a dictionary entry, the tool flags it. This approach catches transpositions ("teh" instead of "the"), missing characters, doubled characters, and phonetic misspellings where the writer knew the word but reproduced its spelling incorrectly.
These are errors of mechanical reproduction: the writer knew the word, had the right word in mind, but the encoding from thought to keystrokes went wrong somewhere. Spellcheck corrects the encoding error.
For native speakers, this is often the majority of the writing errors they make in the normal course of professional work. They know the words. They sometimes type them wrong. Spellcheck fixes this automatically and they move on.
The errors that spellcheck cannot catch
Non-native writers make a different kind of error, one that spellcheck has no mechanism to detect. They write correctly spelled words that are the wrong words for the sentence.
Article errors are the clearest example. "I had conversation with manager about the project" uses only correctly spelled words. There is no misspelling. Spellcheck produces no flag. But "a conversation" is the correct form, and "the manager" or "my manager" depending on context. The error is grammatical and contextual, not orthographic.
Preposition errors follow the same pattern. "We discussed about the proposal" is all correctly spelled words. "Discussed" does not take "about" as a direct companion in standard English, but spellcheck has no way to know this. It sees "discussed" and "about" as correctly spelled words in proximity and produces no alert.
Subject-verb agreement errors in complex sentences are another category. "The team of developers have decided" passes spellcheck entirely. The grammatical subject "team" (singular) disagrees with "have" (plural), but both are correctly spelled and spellcheck does not parse sentence structure.
Register errors do not even register as a category in spellcheck's model. "Kindly be informed that I will attend the meeting" is grammatically correct and fully spellchecked. It is also significantly more formal than contemporary professional email norms in most Western workplaces, which means a native speaker reader will likely notice a register mismatch. Spellcheck cannot see this because spellcheck does not model register.
Why these gaps exist
Spellcheck was not built to catch these errors because the tool was designed for a population where these errors were not the main problem. Native English speakers in professional contexts do not typically make article errors, preposition errors of the kind described here, or register errors. They may use passive voice too heavily or write overly long sentences, but these are style problems rather than the structural interference patterns that come from writing in a second language.
The populations that built the earliest writing tools and the populations those tools were designed to serve were largely the same: native English speakers in educational and professional environments. The tool fit the population. For a different population, one where the dominant error patterns are about language acquisition and L1 interference rather than mechanical typing errors, the tool does not fit as well.
What actually helps non-native professional writers
The gap that spellcheck leaves is a grammar and style gap, not a spelling gap. Catching it requires tools that parse sentence structure and understand grammatical rules, not just tools that compare words against a dictionary.
Grammar checkers do this better than spellcheck, but they are not all equally useful for non-native writers. A grammar checker trained on academic prose will model academic norms. A grammar checker trained on professional email will model professional email norms. For a non-native writer trying to improve their workplace communication, the latter is considerably more relevant.
The register gap is harder to address. Modeling what "appropriate register for this context" means requires understanding the context, the relationship between writer and recipient, and current norms in the relevant professional culture. This is where the most interesting work in writing assistance is happening, and where simple rule-based tools reach their limits.
What Linguix focuses on
We built Linguix specifically because we saw that spellcheck and basic grammar tools were leaving the most common and most consequential errors in non-native professional writing unaddressed. Our focus has been on the categories of error that matter most in the professional email and document context: article and preposition errors, subject-verb agreement at distance, passive constructions where active voice would be clearer, and register signals that suggest a mismatch between the writer's default and the context they are writing in.
We are not building a comprehensive grammar reference or a second-language acquisition platform. We are building a tool that catches the specific patterns that non-native professional writers actually encounter in the course of their working days, inline in the applications they are already using. Spellcheck is useful and we do not compete with it. We start where spellcheck stops.