Suelto is a Spanish learning system for people who already know more Spanish than they can use. If you live in Spain, understand most of what you hear, but still freeze, simplify, or dodge when it’s your turn to talk, Suelto was built for that specific problem.
The curriculum itself runs the whole way, A1 to the end of C1, so a complete beginner can start at lesson one and never leave the path. But the design centre sits further along. Suelto assumes the interesting problem is the distance between what you recognise and what you can actually produce under pressure, and everything is built to measure that distance every day. If you already have a foundation, the placement drops you at it and the machinery is at its sharpest.
The short answer
Suelto measures the gap between what you recognise and what you can produce, and rebuilds tomorrow around today’s specific failures. That is an unusual thing to build, because the interesting signal at this stage is not in your errors.
Batia Laufer’s longitudinal work on passive and active vocabulary found the ratio of active to passive size falling from 89% to 73% across a year of instruction: learners improving throughout, comprehension outrunning production, the gap widening rather than closing. A system that only scores what you attempted cannot see that happening. Everything below is an attempt to see it.
Five steps, thirty minutes a day
Every day in Suelto follows the same five-step loop, in the same order: Learn, Lesson, Conjugation, Speak, Review. The order is not arbitrary. Each step feeds the next, and skipping one breaks the chain.
Learn introduces today’s material: a grammar point, a set of vocabulary items, or a functional chunk (how to hedge, how to refuse politely, how to express doubt). This takes about five minutes and focuses on meaning first, forms second. You see examples drawn from real Peninsular Spanish, not textbook sentences.
Lesson puts the material into structured exercises: fill-ins, transformations, sentence building. This is guided production with immediate feedback. You’re not reading about Spanish; you’re writing it. Ten minutes, timed, with corrections that explain the reasoning.
Conjugation is a separate, dedicated step. One verb, one tense, all six persons, timed individually. You need three clean passes under a speed threshold before it counts. This step exists because conjugation is a retrieval problem, not a knowledge problem, and it needs its own pressure.
Speak is the core. You have a conversation with Suelto’s AI counterpart: a structured exchange on a realistic scenario (a work problem, a weekend plan, a bureaucratic appointment). You write your responses, then read them aloud. The system analyses both what you said and what you avoided.
Review consolidates. Items that wobbled during the day get scheduled for tomorrow. Items that were smooth get pushed further out. The spacing is per-item and per-state, because recognising a word and producing it under pressure are different skills on different schedules.
The five-step shape is not arbitrary either. Paul Nation’s four strands put meaning-focused input, meaning-focused output, language-focused study and fluency development at roughly a quarter each, and the plateau is what a diet of one and a half strands looks like after two years. And the review step is spaced rather than massed because Nicholas Cepeda and colleagues synthesised 839 assessments of distributed practice across 317 experiments and found spacing reliably beating massed repetition, with the optimal gap widening as the retention interval grows.
You produce first, then the system listens
Most language tools follow a teach-then-test pattern. You read a grammar explanation, then you do exercises that practise the thing you just read. The exercises are scored right or wrong. You move on. The problem with this loop is that the exercise always tells you what to produce. You never have to decide, under pressure, which structure to use. That decision is the skill that conversation requires, and it’s the skill most tools never train.
Suelto inverts this. The conversation step puts you in a realistic scenario and asks you to respond before the system has told you what to say. You write your answer. The system reads it. Then it shows you what happened: where you were accurate, where you made errors, and (this is the part that matters) where you took a simpler path than you needed to.
The conversation is where most of the diagnostic data gets generated. Every response you write is a sample of your actual Spanish under mild pressure, not a test, not a fill-in-the-blank, but an open production task where you choose your own structures, vocabulary, and register. That sample goes into your profile and shapes tomorrow’s session.
The corrections are specific. If you wrote something redundant, you see why. If you used a construction that works but isn’t what a native in Spain would say, you see the Peninsular alternative. Every correction is something you can read aloud and practise immediately, so the feedback closes in the same session it was generated. For a deeper look at why this production-first approach matters, see why you understand more than you can use and the gap between passive and active vocabulary.
What you avoid matters more than what you get wrong
Traditional scoring looks at errors: wrong verb form, bad agreement, incorrect preposition. Suelto looks at those too. But the more revealing signal at the intermediate level is not what you get wrong. It’s what you never attempt.
When you say pero instead of aunque, that’s not an error. Both words are correct. But if you always say pero and never say aunque, that’s a pattern. When you say creo que no es buena idea instead of no creo que sea buena idea, you’ve dodged the subjunctive by restructuring the sentence. Grammatically fine. Diagnostically significant.
Suelto tracks these avoidance patterns across sessions. It compares what you said to what was available at your level, the structures and vocabulary you’ve been taught but consistently don’t use. When a pattern appears (three sessions in a row where you ducked the subjunctive after a negative belief verb), the system schedules a targeted exercise. Not more grammar explanation. A production task where the meaning you want to express requires exactly the structure you’ve been avoiding.
This is the most important difference between Suelto and a tool that only scores right and wrong. A correct sentence can still carry diagnostic signal. Most tools will never flag creo que no es buena idea because it contains no error. Suelto flags it because your profile says you’ve been taught the subjunctive after negative belief verbs and you’re consistently not using it. The error isn’t in the sentence. The error is in the pattern across sessions.
None of this is our idea. Elaine Schachter named the phenomenon in An Error in Error Analysis in 1974: her learners were not getting English relative clauses wrong, they were producing almost none, so counting errors ranked them as the strongest group and counting attempts reversed the ranking. Martin Kleinmann’s follow-up supplied the mechanism, finding that what predicts a detour is confidence in that one specific structure rather than general ability.
Fifty years on, essentially no consumer learning tool records it, because recording it requires knowing what was available and unused rather than only what was submitted.
Your avoidance is the map of your actual gaps. For more on this, see subjunctive avoidance and why the B1 plateau is a measurement problem.
A word is not “known” until it comes out under pressure
Most apps track vocabulary as a binary: known or not known. You identified imprescindible in a multiple-choice exercise, so it’s marked learned. Green check. Progress bar fills.
But that recognition test only proved you can spot it when you see it. Suelto tracks five distinct states for every vocabulary item: Recognised (you understood it in context), Retrieved (you produced it when given a cue), Produced (you wrote it in a sentence without prompting), Spoken (you said it aloud at conversational speed), and Spontaneous (it appeared unprompted in your free production).
Each state requires different practice and has its own spacing schedule. Moving from Retrieved to Produced is a different challenge than moving from Produced to Spoken, and both are harder than moving from unrecognised to Recognised. The system tracks where each item sits on this ladder, and the daily session targets the specific transitions that are due.
The ladder is borrowed rather than invented. Batia Laufer and Zvia Goldstein built a test around four degrees of knowing a word and validated it on 435 learners in Testing Vocabulary Knowledge, finding the hierarchy held at every word-frequency level and, more usefully, that the harder rungs predicted real classroom performance better than the easy one. The thing cheap tests measure is the thing that matters least. We add a fifth state above theirs, spontaneous use, because that is the only one that counts in a conversation.
Your vocabulary is not one number. It’s five numbers, one for each state, and the distance between the first and the last is the gap you feel every time you open your mouth. See passive vs active vocabulary for the full picture, and why conjugation tables don’t build the production path.
One verb per day, timed until automatic
The Conjugation step gives you one verb in one tense. All six persons. Timed individually. Your goal is not to get them right (you probably can already). Your goal is to get them right fast, under two seconds per form, three clean rounds in a row.
Speed matters because conversation doesn’t wait. If you need four seconds to assemble tuvieras, the moment is gone. The person you’re talking to has already moved on, or you’ve restructured your sentence to avoid the form entirely. Suelto treats latency as a first-class metric. A correct answer at four seconds is not the same as a correct answer at one second, and the system tracks the difference.
Norman Segalowitz separates three kinds of fluency, and the binding one here is cognitive fluency: the efficiency of the machinery that reaches a form and assembles an utterance. It is the variable an untimed exercise cannot observe, which is why a learner can score full marks on material they will never produce in time. Robert DeKeyser’s skill acquisition account explains the route: knowledge moves from declarative through procedural to automatic, and what automatises is whatever you actually practised under the conditions you need it in.
And the reason the step is retrieval rather than review is Roediger and Karpicke’s test-enhanced learning work, which found restudying winning on an immediate test and losing badly on a delayed one. Looking at a paradigm again feels productive and is not the learning event.
Verbs are gated to your lesson progress. You never drill a tense you haven’t been taught. And if you live in Spain, every drill includes vosotros, because that sixth form is the one most global apps skip and the one you actually need in every casual conversation. For more on why timed retrieval works where tables don’t, see conjugation tables don’t make you conjugate.
What Suelto is not, and when something else is better
We would rather say this here than have you discover it in week three.
A tutor is better if you can afford one and find a good one. A skilled teacher is a measurement device with judgement: they hear the sentence you did not say, ask the follow-up you were hoping to avoid, and refuse the version of your answer that was merely acceptable. That is the same job Suelto is doing, done by something that also understands you. The catch is cost, scheduling, and the fact that many tutors do not systematically probe avoidance either, because it is hard and the lesson is more pleasant without it.
A comprehensible input platform is better if your comprehension is genuinely weak. If you cannot follow a conversation at all, your bottleneck is not production and a diagnostic aimed at production will tell you things you cannot yet act on. Go and build comprehension, in volume, and come back.
Nothing here replaces living the language. Input, media, friends, arguments, boredom in Spanish: that is where the language actually goes in. Suelto is a way of aiming a portion of your week, not a substitute for the rest of it, and a learner who does thirty minutes here and nothing else will improve slowly.
And we are not an exam course. If you need DELE B2 in eight weeks, buy past papers and a teacher who marks them.
What we think Suelto is genuinely best at is narrow: seeing the specific structures you avoid, keeping recognition and production on separate books, and rebuilding tomorrow around today’s failures. If that is not your problem, it is the wrong tool.
Built for Spain
Suelto teaches Peninsular Spanish. Not neutral Spanish, not Latin American Spanish, but the Spanish spoken in Spain: vosotros forms in every tense, present perfect as the default recent past, vale and quedar and mola in the vocabulary, and scenarios drawn from life in Spain.
This means your conversation practice includes situations like scheduling a cita previa, explaining a problem to your landlord, navigating a work meeting where everyone interrupts, and understanding what the person at the ayuntamiento just asked you. The administrative register, the social register, the informal register at a dinner with friends: all Peninsular, all practiced as production, not just comprehension.
If you learned Spanish from a global app or a LATAM-focused course and then moved to Spain, the mismatch is real. Suelto doesn’t ask you to relearn Spanish. It recalibrates the Spanish you have for where you actually live, including the administrative vocabulary that no global course covers.
The goal is not to know more Spanish. It’s to make the Spanish you already know available when you need it: at the pharmacy, on the phone, at the table with friends, in the meeting where everyone is talking at once. That’s the problem, and Suelto is the system that solves it.