Why your tutor should remember you
The stranger problem
Imagine booking a human tutor who, at the start of every lesson, has no memory of you. They do not know your level, so the first ten minutes go on finding it. They do not know you are learning English to change jobs, so they pick a topic at random. They do not know that you have confused the present perfect with the past simple in every one of the last four sessions, so they correct it once, gently, as if it were new — and then never follow up.
That tutor could be excellent in the room and still be nearly useless over a course. Not because their teaching is bad, but because correction only works across time. A mistake pointed out once and never revisited is a mistake you will make again. A level estimated fresh each week is a level nobody is tracking.
A conversation partner with no persistent record is that stranger by construction, however good the conversation itself is. The relationship simply does not exist between sessions. And because the conversation is the visible part, the missing half is easy not to notice until several months have passed with no measurable change.
What has to be remembered, specifically
“Memory” is a vague word for what is actually a short, boring list. A tutor that continues a course rather than restarting one needs to carry a handful of things into every session, and each of them changes what happens in the room.
Your level, with a confidence rating. Not a badge — a working estimate that drives how hard the tutor pitches the conversation, how fast it speaks, and how quickly the plan moves. An estimate held with low confidence should behave differently from one held with high confidence: it should be probed more.
Your goal and your timeline. Learning English for a relocation, for an exam, for work, or for travel produces genuinely different courses. The goal decides the vocabulary domains, the scenarios, and what “ready” means. The timeline decides pacing.
Your interests. This sounds like a nicety and is not. Interests decide what the practice is about, and material a learner finds boring is material they disengage from early. The same grammar target can be practiced through a conversation about football or about supply chains, and one of those will get more words out of you.
Your error patterns. Which mistakes recur, with examples in your own words and a rough frequency. This is the single most valuable thing a tutor can hold about you, and it is the thing a fresh session structurally cannot have.
Your vocabulary state. What you are learning, and what you have shown you know. Without it, a tutor either re-teaches words you have, or uses words you do not.
Your progress against the plan. Which curriculum objectives are being introduced, which are being practiced, which are done. This is what stops a course from circling.
Your homework results. What you did between lessons and, more usefully, what you got wrong in it.
The parameters that are not about English at all
There is a second, smaller set of things worth remembering that have nothing to do with language and quite a lot to do with whether a lesson is bearable.
How much support you need in your own language, which is high at the beginning and should decay as your level rises. How fast the tutor should speak, which is a per-learner setting rather than a global default, and which a learner should be able to change mid-conversation just by asking. And how long the tutor waits before deciding you have finished your turn.
That last one belongs on the list for the same reason as the others: the right amount of patience is different for different people, and different for the same person at A2 and at B2. Whether it is remembered or re-guessed every session decides how a lesson feels long before it decides anything about your English. What it does to the balance of a conversation is a subject in its own right, and one worth separating from this list — the point here is only that it is a fact about you, and belongs wherever the rest of the facts about you live.
One teacher, many characters
There is an obvious tension between continuity and variety. Practice needs situations — a job interview, a doctor visit, a hotel check-in, a salary negotiation — and each of those needs the tutor to play a character. A hiring manager, a receptionist, a budget-conscious manager. If the tutor is a different entity in each scenario, the relationship fragments and you are back to talking to strangers.
The resolution is that the character sits inside the scenario, and the teacher sits outside it. The same tutor, with the same name, tone and memory, runs the role-play and then steps out of it to correct, explain, or close the lesson. You are not meeting a new persona each week; you are watching your teacher act, and then be your teacher again.
That distinction also decides who owns the feedback. A character in a role-play cannot tell you that you have now made the same article error in four sessions, because a receptionist would not know. A teacher can, and should.
What memory changes in the room
The concrete effect is that a lesson can start in the middle. There is no re-establishing of level, no “so what would you like to work on today,” no polite re-introduction. The tutor already has the plan for this session, the review items due today, a short summary of how the last conversation went, and the results of whatever homework happened in between.
It also changes what correction can be. A tutor with no memory can only correct in the moment, which means every correction is equally weighted — a slip and an entrenched habit get the same treatment. A tutor with a record can rank them, which is the harder and better behaviour: let the slip go, and spend the correction on the pattern that has now appeared four sessions running. The ranking is the thing memory buys. Without a record there is no way to know which of two identical-looking mistakes is the one worth interrupting for, so either both get interrupted or neither does.
And it changes what a review looks like. Instead of a drill announced as a drill, a due item can be woven into the conversation as an ordinary opportunity to use the structure again. You do not experience it as a test. You experience it as the topic happening to require the tense you got wrong last week — which, from the tutor’s side, is not a coincidence at all.
Memory is a design decision, not a feature
It is worth being clear that none of this is emergent. A conversational model does not remember you because it is intelligent; it remembers you because something explicitly stored what happened and explicitly loaded it back in before the next conversation started. Each dimension above lives somewhere specific, gets updated by a specific step, and gets read at a specific moment. Skip the storing step and the most capable model available is still a stranger every Tuesday.
That is also why memory and the closed learning loop are the same subject viewed from two sides. The loop is the mechanism that writes; memory is what it wrote. Neither is useful alone: writing a record nobody reads is bookkeeping for its own sake, and a tutor that would happily read a record has nothing to read if no step ever produced one.
The test you can run yourself
Ask a tutor, in your fourth or fifth lesson, what you keep getting wrong. A tutor with a record will name a specific pattern and probably give you your own sentence back. A tutor without one will produce something plausible and general — a description of what learners at your level typically struggle with, which is not the same thing and is much less useful.
The method page sets out how the curriculum sits around this, and how it works walks through what actually happens from the intake conversation onward.