Lexum
From translation to real vocabulary
For most of Lexum's development, I've been working with a very small testing circle.
That was enough while I was still changing fundamental parts of the product every few days.
But I'm starting to notice the downside.
I know where everything is.
I know what every button is supposed to do.
I know why certain decisions were made.
Which also means I'm not the right person to judge whether any of it is obvious to someone seeing Lexum for the first time.
The product has also expanded enough that I want to see how it behaves across more native language → learning language pairs.
So I'm starting to bring more people into the pre-beta.
I'm not particularly interested in "looks good" feedback at this stage.
The useful stuff is more like:
"I didn't understand what this was for."
"I expected this to happen instead."
"I wouldn't use this."
"This part was actually useful."
"I saved a few words and then had no idea what to do next."
That's the kind of feedback I'm looking for now.
If you're learning a language and wouldn't mind trying an early build, I'd love to know:
your native language + the language you're learning
Lexum — pre-beta
lexum.app
One thing I've been thinking about while building Lexum is where learning material should actually come from.
Many vocabulary tools start with a predefined set of words.
But when I pay attention to my own language use, many of the words I actually want to remember come from something I was already trying to understand - an article, a document, a message, a book, etc.
So I've been building a feature called Breakdown.
The basic idea is simple:
paste real text → identify potentially useful vocabulary → explore the words that matter → keep what you want to learn
What I like about this model is that Lexum isn't deciding the subject first. The user's content is the starting point.
The harder product question is how much the app should interfere.
If Lexum highlights too much, it becomes noisy.
If it highlights too little, it isn't useful.
And level matters too — the vocabulary worth surfacing for an A2 learner isn't necessarily useful for someone at B2.
I'm still tuning that balance.
I'm curious how other people would approach this:
Would you rather have the app automatically choose what looks worth learning, or mostly give you the tools to pick words yourself?
Lexum — pre-beta
lexum.app
I haven't shared a proper Lexum update for a while.
A few months ago the product was still fairly close to a simple loop:
translate → save → repeat
It worked, but while building and using it I kept running into the same problem: storing a translation doesn't necessarily help someone understand or remember the word.
I didn't make one big pivot. Instead, a lot of smaller product decisions gradually changed the direction.
The mental model I'm using now is closer to:
encounter → understand → keep → practise → remember → expand
The original translation flow is still there, but I'm trying to make the things around it feel like one connected learning process rather than a collection of separate features.
A lot has changed in the product over the last few months, so instead of dumping a changelog here I'm going to share some of the individual decisions and experiments separately.
I'm also reaching the point where I need more people outside my tiny testing circle using it.
Lexum is still pre-beta and definitely has rough edges, but that's exactly the kind of feedback I need now.
If you're learning a language and are interested in testing it, I'd love to know your native language + the language you're learning.
One thing I slowly realized while building the app:
Translation alone is useful, but it still keeps words somewhat isolated.
Adding pronunciation, example sentences, and common word pairings changes the feeling quite a bit. Words start to feel more connected to actual usage rather than just "saved vocabulary".
It's still the same core idea:
translate → save → revisit
But with a bit more context around every word.
Interesting how relatively small UX/content additions can noticeably change retention and overall experience.
Curious if others building language or learning-related products noticed something similar.
While learning languages, I noticed something strange.
I was translating a lot of words every day, but remembering almost none of them.
Translators helped me understand texts in the moment, but they didn’t really help me build vocabulary.
Another problem is that new words appear everywhere — articles, chats, videos, and conversations.
Often, there’s no good moment to save them, and even if you do, they end up scattered across notes or different apps.
Learning becomes a bit chaotic.
So I started building a small tool called Lexum to solve this for myself.
The idea is simple: translate the starting point for actually learning and remembering words.
The app is currently in pre-beta, and I'm preparing the first test group.
Curious — how do you usually save or remember new words when learning a language?
-
2
The shift from “translate → save → repeat” to a connected learning process makes sense. I’ve found that saved vocabulary can easily become a graveyard of words you intended to revisit but never actually did.
Curious what part of the new flow people testing Lexum have found most useful so far.
-
1
Thanks - “a graveyard of words” is actually a very accurate description 😄
It’s still too early to call a clear winner from tester feedback, to be honest.
What seems most useful so far is simply closing the gap between “I saved this word” and “I actually did something with it” - getting it back into practice, with enough context that it doesn’t feel like a random flashcard.
I’m starting to bring more testers in now, so I’m really curious to see whether that holds up with a larger group.
-
1
That makes sense. The interesting thing to watch now is whether that “saved → actually used” gap is consistently the part testers value, or whether something else emerges once you have more people using it.
-
-
1 Comment
The distinction between “looks good” and “I didn't understand what this was for” is important. Curious what the first-time testers end up noticing that you completely stopped seeing yourself.