You have read the explanation of the Spanish subjunctive many times over your language learning journey, in several places. You could explain the trigger conditions to a friend over coffee and get them right. Then you are in a conversation, you reach the point in the sentence where the mood has to change, and you freeze. You produce the indicative, or you find another way to say what you wanted to say.

Nothing’s wrong with your knowledge. What you are running into is that studying a structure and producing it are two different activities, and doing the first one well often is not enough. Most learners have never been told that directly, because most language products are built in a way that implies the opposite: read, then answer some questions, and the ability to speak is supposed to magically arrive.

Dioma is built the other way around. It is a practice engine. It exists so that you produce language and get told precisely what was wrong with what you produced. That is a narrow choice with real tradeoffs, and it is the choice that decided everything about how instruction works inside the product, including the Learn screen that sits inside of every speaking and writing exercise.

Knowing the rule and reaching for it are two different jobs

Research on how skills get built has long treated knowing about something and being able to do it as two different achievements. Speaking under pressure runs on habits you can only build by speaking.

This is why recognition tasks and production tasks are not interchangeable. Picking the correct conjugation out of four options is recognition: the right answer is in front of you and the job is to spot it. Generating that conjugation from nothing, inside a sentence you are also assembling, while a person waits, is an entirely different scenario. A learner can pass the first reliably and fail the second reliably, at the same moment, on the same structure.

Compare it to driving. You can know every rule in the highway code and answer any question about right of way. That knowledge does not give you the ability to merge: accelerating into a gap you have to judge, with a truck behind you, while watching the exit sign. The rules are necessary, and the merging is a separate thing you build by merging.

Producing language is not a test of what you learned. It is the activity that turns what you learned into something you can use.

There is a second cut inside that one, which Dioma has written about separately: not all producing counts as practice either, and the kind of practice matters more than the amount.

Why most language products blur the two

Look at the shape of a typical lesson: an explanation, then exercises that check whether you absorbed it. Multiple choice. Match the pair. Tap the words in order. Fill the blank from a word bank. Almost all of it is recognition, and almost all of it is presented as practice. That is not a stupid design, and it is the right one for beginners, who need a base of recognition before they can produce anything.

It stops working at the point where you already recognize far more than you can produce. We are really fascinated by this area and grounded our entire methodology in it. Dioma has written about many elements of this gap: how input keeps building understanding without building speech, what the acquisition research says about how adults get there, and why beginner apps stall around B1 once tapping the right answer stops being the hard part. The consequence for product design is the one this piece is about. If the learner’s gap is production, the product has to make you produce, not just study.

What Dioma gave up by choosing to practice over teaching

Dioma is not the place you learn a language from zero. There is no Beginner curriculum and we are not planning to add one anytime soon. If you cannot yet produce sentences in the language, there is nothing for Dioma to correct.

A course can promise to take you from nothing to somewhere. Dioma requires that you show up with a base and a goal, which is a harder sell, and which filters out anyone looking for five minutes a day of learning.

Going deep means going narrow. Dioma intentionally started with three languages, Spanish, French, and Hebrew, rather than thirty, because a practice engine needs a full expansive topic set per level to take you through, and every one of those topics is written by a person. Because of this, we decided to not have a free tier. Anything given away at scale has to be simple to serve and monetize your usage in some way, we decided to deliver real practice, backed by a real curriculum and charge an affordable price to access it.

Lastly, a practice engine is harder to demonstrate than a lesson. A lesson looks like progress immediately. Practice looks like work, and you only see what it did after a few weeks of it.

When instruction should appear, and who decides

None of that means Dioma refuses to teach. For adult learners, being handed the rule is an important part of learning. Norris and Ortega’s 2000 analysis of instructed second-language acquisition found that explicit teaching of rules outperforms leaving adults to infer them from exposure. We do not learn language via osmosis.

Our first version had no explicit teaching, and in talking to all of you, we know that was not the right choice.

So, as a practice-focused engine, we needed to decide where to have learners learn, and where to have them practice.

Dioma’s answer is that the learner decides, at the moment they notice they need it, without leaving the exercise. Every speaking and writing exercise carries a button labelled “Learn More”, sitting beside the name of the topic you are working on. Tap it and the lesson for that topic opens over the exercise under the heading “Learn the topic”: the written explanation, worked examples you can hear read aloud in the target language, the errors learners make on that structure with the wrong version beside the right one, and a short check to see whether the explanation landed. Close it and you are back at the same prompt, ready to answer.

It never opens itself and there is no setting that makes it appear automatically. That is a deliberate choice to give you as the learner control over if you’d like to review the material or jump right in.

Why the quiz inside it is not scored and never recorded

The Learn screen ends with a short fill-the-blank exercise. Tap a choice and you get an instant verdict. No submit button, no tally, no percentage, no celebration.

The result is also not written down anywhere. It does not enter your session history, and it cannot move your level, your practice plan, or the record of which topics you have covered. A fill-the-blank check is a recognition item, and the record Dioma builds about you is assembled from production: what you said and wrote, and what the corrections found in it. Letting a recognition result into that record would quietly corrupt the one measurement the product exists to make.

Why the explanation is the curriculum author’s words, unchanged

The teaching in that overlay is not written when you open it. Dioma’s curriculum team wrote it alongside the topic itself, long before this screen existed, and it was already living inside the app with nowhere for a learner to see it. The curriculum runs more than 700 topics across Spanish, French, and Hebrew, each with its own examples and recorded learner errors, and nearly all of them with a written explanation.

What the overlay does is render that text as the author wrote it. Nothing rewrites, shortens, splits, or summarizes it on the way to your screen, and nothing improvises a fresh explanation in the moment. You can toggle between English and your target language.

The people who wrote it had designed CEFR-aligned curricula before and had taught adults one on one, and they chose these examples and recorded these particular errors because they had watched learners make them.

Frequently asked questions