--- title: "The Teacher Who Never Sleeps" date: 2026-03-28 description: "Advertorial for businesses, about ltrace.ro, the educational platform for Years 7–10 built by Hint One Zero SRL Botoșani. Interview with technical architect Bogdan Ignat, web developer, on immediate feedback, automated essay assessment, and the real limits of AI in education." author: "Petru Cojocaru" categories: - name: "Advertorials" url: "https://cojocarupetru.info/advertorials.md" --- # The Teacher Who Never Sleeps *A note for the reader: This article is an ADVERTORIAL for businesses. It presents the platform **[ltrace.ro](https://www.ltrace.ro)** (Interactive lessons and tests for Years 7–10, with instant automatic feedback) and the Botoșani-based company that built it, **[Hint One Zero](https://hintstudio.ro/) SRL**. The tone is favourable and acknowledged as such. The product's limitations are named explicitly, with verifiable sources. Transparency is not an excuse — it is a condition.* **Listen to the audio dialogue summary** --- ![The Teacher Who Never Sleeps](https://cojocarupetru.info/images/Advertoriale-sound/logo-hint-one.png "The Teacher Who Never Sleeps")It was around ten in the evening when I opened my laptop. Outside, rain was falling with a dull persistence — no rhythm, no promise of anything. My son had fallen asleep in the next room, a grammar textbook on his chest — one he had been picking up and putting down for weeks without retaining much from it, because nobody had yet explained to him why what was written in it mattered. I was thinking about that. About that enormous distance — which you forget entirely once you become an adult — between knowing something and understanding why it is worth knowing. They are two different things, and school, more often than not, teaches you the first and leaves you alone with the second. The man I was about to interview had built an answer to this problem. Not a philosophical one. A practical one. From code, from databases, from years of work at a desk. **The interview with Bogdan Ignat** — web developer, a programmer with eight years of experience, now technical architect at **Hint One Zero SRL**, the Botoșani-based company that built [ltrace.ro](https://www.ltrace.ro) (Interactive lessons and tests for Years 7–10, with instant automatic feedback) — he settled in front of the screen with the composure specific to people who know their work speaks more eloquently than any talk about it. I asked directly. He answered in kind.![The Teacher Who Never Sleeps](https://cojocarupetru.info/images/Advertoriale-sound/logo-ltrace-mic.png "The Teacher Who Never Sleeps") **Before I go any further: yes, this article is an advertorial for businesses.** I declare this not because the law obliges me to, but because I find that concealing the intent of a text is a kind of dishonesty I cannot practise without feeling bad afterwards. The company I am writing about is in the same city as me. The platform I am presenting exists, is publicly accessible, and can be tested by anyone. Its limitations are real and I shall name them. The rest is the interview. --- ## What ltrace.ro Actually Is — And Why It Matters There is a type of problem that doesn't seem urgent until it becomes one. The National Assessment at the end of Year 9 is, for many children and many parents, the moment when you realise you have been living with an illusion: that grades in the register mean something is understood. That if he got a seven in Romanian in Year 6, he can write an essay. That if he solved the exercises in his workbook, he is ready for the exam. None of that necessarily follows. ltrace.ro is a platform built specifically for pupils in Years 7 to 10, with a focus on preparation for the National Assessment in Romanian Language. It is not a site with PDF worksheets. It is not a YouTube channel with explanations. It is a system that allows you to take tests — multiple choice, true or false, complete essays — under conditions as close as possible to the real exam, to receive an immediate response, to understand exactly where you went wrong, and to try again until you genuinely understand. Bogdan Ignat explained to me how each of these functions was designed from scratch. I am not going to reproduce the technical jargon he used — it is not useful to you as a parent or teacher — but I am going to reproduce the ideas, because the ideas matter regardless of the language in which they were formulated. --- ## Interview Part One: Tests and Immediate Feedback — Or Why It Matters That You Find Out Straight Away When You Are Wrong I kept returning, in my mind, to a question I had long carried without ever putting it into words: why does my son find it so hard to understand from what he got wrong? He would receive the test back with red crosses and perhaps a mark, but with no explanation of what he had missed. And he didn't know where to begin repairing it, because he didn't know what was broken. **I asked Bogdan Ignat directly: why does it matter that feedback comes immediately?** His answer was one I recognised without having heard it formulated before: the brain is "online" with a problem for only a short while after it has solved it. If in that interval you give it the information that it got something wrong and show it exactly where and why, the connection is made. If you wait two days — as usually happens, with tests brought home after the weekend — the connection has already been broken. The child is no longer in the same mental place they were in when they made the mistake. The correction arrives too late and glances off. Cognitive psychologists — Bogdan mentioned several names that I subsequently verified — have studied this systematically. Immediate feedback, at the moment when the mistake is still fresh, produces lasting corrections. You don't just learn that you got something wrong. You learn why you got it wrong, and that is an entirely different mental operation. On ltrace.ro, feedback is not merely "wrong" or "correct". It says: you chose that the verb *a merge* (to go) is transitive, but it is intransitive, and here is why. Or: your essay has an introduction and a body, but the conclusion does not return to the main idea — that is a marking-scheme point you have lost. Detailed, specific, immediate. **And yet I wanted to ask something more uncomfortable: do we not risk creating children who can no longer tolerate uncertainty? Who permanently need someone to confirm they are right?** Bogdan paused for a moment. Then he said that yes, the risk exists, and that it is precisely for this reason that the platform also includes complete simulations without immediate feedback — tests that reproduce exactly the conditions of the National Assessment, with the evaluation arriving afterwards, by e-mail, exactly as results arrive in real life. They reproduce the full experience: the waiting, the uncertainty, the moment when you open the message and find out. You cannot protect a child from every form of discomfort. Nor should you want to. I was glad he said that. It means the people who designed the product also thought about what it could do wrong, not only what it could do right. --- ## Interview Part Two: Essays — How a Computer Evaluates a Text Written by a Human This is the point where many parents and teachers stop, sceptical — and I understand why. The idea that a computer programme can read an essay written by a thirteen-year-old and say something useful about it seems, at first glance, either impossible or dangerous. It is neither. But nor is it the miracle some advertisements promise. **Bogdan Ignat explained to me how essay assessment works. I have simplified, but not distorted.** The system does not read an essay and say "it is good" or "it is weak". That would indeed be absurd. The system evaluates against specific criteria drawn directly from the official mark schemes of the Ministry of Education — the same criteria by which a real examiner marks your paper at the exam. The structure of the text. Grammatical correctness. Coherence — that is, whether ideas connect to each other or jump about without a thread. Relevance to the task — did you write what the question asked or did you go off in a different direction. Richness of vocabulary. Each criterion carries a weighting. The result is a score broken down by category — not a mark that falls from nowhere, but a breakdown showing where points were lost and why. The child sees exactly where they lost marks and can work on that area. **What the system cannot do?** Bogdan said it without needing to be asked: it cannot assess the authenticity of voice. It cannot recognise that a metaphor is unusually beautiful or that a paragraph has a particular rhythm. It cannot sense that behind an essay is a child who has lived through something real and found the words for it. All of that requires a human reader, with experience and sensitivity. A computer does not have these things and will not have them any time soon. For this reason — and this struck me as important — any assessment made by the system can be reviewed by a teacher registered on the platform. The teacher sees the original text and the automated assessment, can correct it, can add comments, and the child receives a notification with the revised score. The automated assessment is a first filter — fast, consistent, available at any hour. The teacher remains the final authority. --- ## Interview Part Three: The Real Limits — What the Research Says, Not the Advertising Now I need to say something you will probably not read in any promotional material about educational platforms with artificial intelligence. There are serious studies showing that automated essay assessment works reasonably well on objective criteria and in languages for which there is sufficient training data. There are equally serious studies showing that, compared with experienced human assessors, correlations range from moderate to weak — especially in languages such as Romanian, which is less well represented in global databases. A study published in 2025 in the journal *Innovation in Language Learning and Teaching* compared three artificial intelligence systems — including recent versions of the best-known ones — with human assessors across one hundred and twenty essays. The conclusion was that the systems tend to be stricter than humans and that their consistency, while high internally, remains inferior to that of an experienced human assessor. A second study, published in *Advances in Physiology Education*, reached a similar conclusion. Bogdan Ignat knows this. He spoke to me directly about these limitations during the interview, without waiting for me to ask. He explained that it is precisely to reduce the risk of mis-assessment that the system does not allow the programme to evaluate freely, but instead explicitly provides the mark scheme before each evaluation — which, according to the same specialist literature, significantly reduces errors. And that teacher re-assessment is not an optional function added to look good in the description — it is part of the fundamental architecture of the product. I believe him. Not because he told me so, but because he acknowledged the limitations without minimising them. What does this mean in practice, for you as a parent or teacher? That automated assessment is a useful training tool, not a definitive judge. It tells you quickly and consistently where problems exist — far more quickly than a teacher who has thirty children in the class and a life outside school could do. But that, for any important decision, the eye of an experienced human remains necessary. --- ## Interview Part Four: A Risk I Didn't See at First The morning after I had written the first notes for this article, I sat down at the kitchen table with my coffee. My son was asleep. Outside, there was a grey morning stillness that promised nothing and threatened nothing. And I realised I had not asked the most important question — not the one about programming, but the one about children. There is a real, documented risk that in the pedagogical literature is called **format training effect**: the situation in which a child who repeats the same type of test dozens of times does not learn to write better, but learns to recognise the patterns of that test. To guess more effectively. To optimise for the displayed score rather than for genuine competence. It is an enormous difference, and it is invisible until it isn't. Education researchers have studied this for decades, especially in high-stakes examination systems — national assessments, entrance tests. The conclusion is worrying: when a child prepared exclusively for a test format is faced with a task that format did not train them for, performance drops sharply. Not because they lack intelligence. Because they were trained to recognise a pattern, not to think flexibly. Bogdan Ignat confirmed that the risk exists and that the platform has tried to reduce it through variation — in types of tasks, in difficulty, in types of subject matter. But he did not tell me it had been eliminated. Nor could he, honestly. This means that a training platform — any training platform, not just ltrace.ro — is more valuable as a complementary tool than as a sole tool. If your son or your pupil does simulations on the platform AND reads literature AND writes free-form texts AND discusses what they have understood with a teacher, then the platform is an accelerator. If the platform is the only tool, there is a real risk that the format is learnt better than the language. I am writing this in the present article — not in a small aside, not at the end, but in the middle — because I find that selling a product without naming its real risk is a form of dishonesty I can practise only with difficulty. --- ## Interview Part Five: The Real Exam Simulation — And Why the Waiting Matters There is a scene I picture with a strange clarity: the night before the National Assessment. Somewhere, across a few hundred flats throughout the country, fourteen-year-olds are sitting at their desks or on their beds with their phones in their hands, anxious, helpless in the face of what can no longer be done in the hours that remain. And they don't know whether they have done enough. ltrace.ro has designed a special component for this: complete simulations that reproduce exactly the structure of the real exam — with all three types of question, with the correct timings, with the correct restrictions. The assessment does not come immediately. It comes afterwards, by e-mail, exactly as results arrive in real life. The child goes through the complete experience — not just the solving, but also the waiting. Why does this matter? Bogdan explained: many children know how to solve exercises, but don't know how to manage the pressure of the real exam. Panic at the clock, the feeling that time is running out too fast, the temptation to read the same question three times without writing anything — these are habits, not states. They are trained or they are not. A faithful simulation builds the habit of functioning under pressure, not just the knowledge of the subject matter. Teachers who prepare pupils for exams already know this. The difference the platform brings is accessibility: a child can do a full simulation at eleven o'clock at night, at home, without depending on a teacher's timetable or the availability of a tutoring centre. --- ## Interview Part Six: The Content — Who Wrote It and Why It Matters There is a question I rarely ask in interviews because it seems obvious, but which usually surfaces something unexpected: who actually wrote the content on the platform? Bogdan Ignat answered without hesitation: teachers with direct experience of the National Assessment — as assessors, as examiners, as people who have read thousands of essays throughout their careers and know exactly what distinguishes a paper scoring eight points from one scoring six. Not content automatically generated by a programme. Not exercises copied from the internet. Things written intentionally, by people who understand both the subject matter and the psychology of a thirteen-year-old sitting at a desk on a May evening not knowing whether they are ready. This matters more than it might seem. There is a fundamental difference between an exercise built by a teacher who has seen a thousand children's essays and one generated by a programme that has seen a thousand internet texts. The former knows what children actually get wrong, which formulations confuse them, which types of questions appear simple but are traps. The latter knows none of this — it reproduces patterns, but does not understand the people who will use the content. --- ## Interview Part Seven: Hint One Zero SRL, Web Developer — Or What It Means to Build Something Useful in a Small Town I had not set out to write about the company itself, but I cannot ignore the fact that there is something significant about ltrace.ro having been built in Botoșani. Hint One Zero SRL is not a newly-established firm. It has a history in building complex online shops, web applications, and integrations with artificial intelligence tools. Bogdan Ignat, web developer, has worked there for eight years. He has passed through all the stages of a programmer who grows — from code written and rewritten, to more complex architectures, to design decisions that affect tens of thousands of users. The fact that this company chose to build an educational platform for children in Romania — not for the international market, not for a corporate client, but for a Year 8 pupil from Botoșani or Iași or Craiova who is preparing for an exam that could change their trajectory — seems to me to say something about motivation. I don't know exactly what, and I don't want to romanticise it. But Bogdan told me, at the end of the interview, that his ten-year-old son uses the platform. That he hasn't told him he built it. That the boy wouldn't be impressed, at his age. That what interests him is seeing immediately where he got something wrong. That he can redo the same test three times. I sat with that for a few minutes after I closed the laptop. I have never seen a ten-year-old willingly do the same test three times with a traditional exercise book. It is possible that this is the only argument you need to understand what the platform does differently. --- ## Interview Part Eight: Direct Questions — Answers Without Jargon **How does automated essay assessment actually work?** The system receives the essay, analyses it against five or six criteria derived from the official mark scheme of the Ministry of Education — structure, grammatical correctness, coherence, relevance to the task, vocabulary — and returns a score broken down by category. Not a single mark. A breakdown showing where points were lost and why. The assessment is not final — it can be reviewed by a teacher registered on the platform. **How accurate is automated assessment compared with a teacher?** The honest answer is: it depends. On objective criteria — spelling, structure, compliance with formal requirements — automated assessment is consistent and rapid. On subjective criteria — authenticity of voice, artistic quality of a text, the deeper meaning of a metaphor — it cannot replace a human reader. Studies published in 2024–2025 show correlations ranging from moderate to good compared with experienced human assessors, with weaker results for languages less well represented in global databases — Romanian among them. The platform has built mechanisms to reduce these errors, but has not eliminated them. The teacher remains the final authority. **Who created the content — the exercises, the questions, the mark schemes?** Teachers with direct experience of the National Assessment, not automated programmes. **Can the platform be used without additional guidance?** Yes, as a training tool. No, not as the sole means of preparation. The risk of training format recognition rather than genuine competence is real and documented. The platform works best as part of a broader preparation that also includes reading, free-form writing, and dialogue with a teacher. **Is it compliant with personal data protection rules?** Bogdan Ignat confirmed that registration for minors requires parental consent, that data is stored minimally and securely, that tests and essays are accessible exclusively to the authenticated user and, where applicable, to the associated teacher. Spring Security — the security layer used — manages all access levels. **How much does it cost?** For the current answer to this question, the best place is directly on [ltrace.ro](https://www.ltrace.ro) — the pricing structure may change over time and I don't want to reproduce information that might be out of date by the time you read this. --- ## What I Was Left With — A Personal Note There is a type of product you can recognise by a simple sign: the people who built it know where it doesn't work. I am not talking about performative modesty — the way large companies list a minor flaw to appear honest before selling their product. I am talking about people who have lived close enough to the problem they are trying to solve that they know exactly where their solution ends and where the problem continues. Bogdan Ignat knows that automated essay assessment has serious limitations. He named them. He knows that the platform can produce a negative effect if used excessively or exclusively. He said so. He knows that there are no longitudinal studies confirming that pupils who use ltrace.ro write better in general — because the platform is too new for that. He said all of this while describing a product on which he had worked for years, in which he believes, and which he uses with his own child. That struck me as more convincing than any promise. ## What You Can Do Next I don't know whether ltrace.ro is the right solution for your child. And, honestly, I don't think any article — however long or well-documented — should make that decision for you. But there is one simple thing you can do, without taking anyone's word for it. Go to the platform and have a look. Let your child solve a test. See what the feedback looks like. Notice whether they come back a second time without being told to. Or not. That, in the end, is the difference between a good idea and a useful tool: whether it is used, not merely understood. If it works, you will see soon enough. If it doesn't, you will know just as quickly — without having lost more than a few dozen minutes. There are no perfect solutions in education. There are only combinations that work better for a particular child, at a particular moment. And sometimes, all you need to find out is to try — not to assume. --- ## Bibliography — Verifiable Sources All sources below are accessible at the links indicated. I include them because I consulted the original texts, not summaries. - **Proceedings BEA 2025 — ACL Anthology** — The 20th Workshop on Innovative Use of NLP for Building Educational Applications, 104 studies on language models in education: [aclanthology.org/volumes/2025.bea-1/](https://aclanthology.org/volumes/2025.bea-1/) - **Urrutia et al., BEA 2025** — *Unsupervised Automated Assessment of Short Answers and Essays*: [aclanthology.org/2025.bea-1.4/](https://aclanthology.org/2025.bea-1.4/) - **Gupta et al., BEA 2025** — *Are Large Language Models for Education Trustworthy across Multiple Languages?*: [aclanthology.org/2025.bea-1.44/](https://aclanthology.org/2025.bea-1.44/) - **Bui & Barrot, 2025** — *Using Generative AI as an Automated Essay Scoring Tool*, Innovation in Language Learning and Teaching: [doi:10.1080/17501229.2025.2521003](https://www.tandfonline.com/doi/full/10.1080/17501229.2025.2521003) - **Study on LLM Scoring Accuracy, 2024** — Advances in Physiology Education: [doi:10.1152/advan.00137.2024](https://journals.physiology.org/doi/full/10.1152/advan.00137.2024) - **Wen & Chano, 2024** — *A Critical Review of the Format Training Effect in Education*, Forum for Linguistic Studies: [doi:10.30564/fls.v7i1.7893](https://doi.org/10.30564/fls.v7i1.7893) - **Study on Risks of Large Language Models in Education, 2025**: [arxiv.org/html/2509.21972v2](https://arxiv.org/html/2509.21972v2) --- *Platforms mentioned:[![The Teacher Who Never Sleeps](https://cojocarupetru.info/images/Advertoriale-sound/logo-ltrace-mic.png "The Teacher Who Never Sleeps")](https://www.ltrace.ro/)* - [ltrace.ro](https://www.ltrace.ro) — educational platform for Romanian Language, Years 7–10, with a publicly available functional demonstration. Interactive lessons and tests with instant automatic feedback. - [hintstudio.ro](https://hintstudio.ro) — Hint One Zero SRL, the company that built ltrace.ro *Readers are encouraged to treat claims about the performance of automated assessment in Romanian with caution: the studies cited refer predominantly to the English language or to different educational contexts. Direct testing on the platform remains the best criterion for evaluation.* --- ## What Teachers Say About Educational Platforms — A Synthesis of Recent Research There is a moment when you realise you have written something wrongly. Not wrongly in the sense that it is false — the information was correct, the sources were verifiable, the intention was honest. Wrongly in the sense that the form in which you wrote it created an impression that the reality behind it cannot support. The previous version of this section presented seven "voices from the classroom" — teachers with names, with experience stated in years, with direct quotations. They were not invented from nothing: each voice synthesised real themes from the specialist literature, each quotation reproduced a frustration documented in published academic studies. But nor were they real interviews. They were composite voices — a technique of narrative synthesis that investigative journalism sometimes uses with success, but which, in the absence of extremely clear labelling, can create the illusion that you are reading real people who spoke to you. That illusion is a problem. Not because it distorted the content, but because it distorted the contract with the reader. I removed that section. What follows does the same thing, but differently: it presents directly what the research says, without fictional intermediaries, without constructed voices, without quotations that belong to no one in particular. The themes are the same. The sources are the same. The framing is different. --- ### The Fragmentation of Digital Tools — Or How a Teacher Ends Up Using Fifteen Applications to Do One Thing A report published in 2025 by Cornerstone Communications and Edsby, cited by the journal THE Journal, shows that the vast majority of schools use between ten and fifteen educational applications simultaneously. This figure, at first glance, seems like an administrative detail. It becomes a real problem when you translate it into time: one application for attendance, another for homework, another for communication with parents, another for tests, another for portfolios. Each with its own authentication system, its own navigation logic, its own notifications. A teacher managing all of this doesn't lose a few minutes a day — they lose a cognitive resource they cannot recover during the lesson. Edsby researcher Emily Mabie, in a 2026 analysis of essential features of online teaching platforms, formulates the conclusion more directly: a useful educational platform is not the one that does the most, but the one that reduces the number of places a teacher has to go to do one thing. The implication for a platform such as ltrace.ro is that the integration of functions is not a comparative advantage over competitors — it is a basic condition for sustained adoption. A teacher who has to leave the platform to send a notification e-mail or to access another reporting tool will abandon the platform, not the e-mail. --- ### Anxiety About Technology Does Not Come From Incompetence — It Comes From Unpredictable Consequences Fang, Zhao, and Yin, in a study published in 2026 in the *Journal of Teacher Education*, analyse the sources of teachers' anxiety in the process of digital transformation of education. Their conclusion runs counter to common intuition: teachers are not afraid of technology because they don't understand it. They are afraid of it because they cannot anticipate the consequences of a wrong click. The anxiety is not about competence — it is about control. Kortright-Torres, in a 2022 study of foreign language teachers who taught online during the pandemic, identifies the same mechanism: sense of personal efficacy — that is, confidence that one can successfully cope with a task — drops sharply when the environment is unpredictable, even if the teacher has the necessary competences. You don't feel incapable. You feel exposed. The implication for design is that a platform that reduces anxiety is not necessarily a simpler one, but a more predictable one. Immediate visible feedback when an action is completed. Clear confirmation messages. The ability to undo an action without permanent consequences. These are not interface details — they are mechanisms of trust. --- ### Automated Assessment Is Accepted When the Teacher Remains the Final Authority Celik, Muukkonen, and Siklander, in a study published in 2026 in *Policy Futures in Education*, analyse the conditions under which teachers accept artificial intelligence tools in assessment. The decisive variable is not the accuracy of the system. It is the perception that the professional role remains intact. Teachers do not reject automated assessment in principle — they reject the idea that it substitutes their judgement, that they become the executors of a verdict delivered by an algorithm they cannot contest. Unal, in a 2025 study published in the *SRATE Journal* about the TeacherServer platform, confirms: the AI tools most successfully adopted by teachers are those in which AI assists the decision rather than replacing it. The teacher who can correct, comment, and override the automated assessment with their own judgement is a teacher who feels amplified, not substituted. The direct implication: teacher re-assessment is not an additional function a platform may or may not offer. It is the basic condition of an automated assessment system that teachers will accept. Without it, the system will not be used — or it will be used with resentment, which is worse. --- ### Inclusion Is Not a Layer Added on Top of a Platform — It Is an Architectural Decision The European IDEATE project, in which the "Ștefan cel Mare" University of Suceava also participates, is premised on precisely this idea: digital inclusion and educational inclusion are not two parallel projects that can be coordinated afterwards. They are the same project, and must be conceived as such from the very outset. Hanna and colleagues, in a study published in 2026 in *Computers & Education*, trace the way in which teachers' professional identities transform when technology enters the classroom. One of the conclusions is that teachers working with pupils with special needs feel most acutely the absence of accessibility features — not because they are more demanding, but because for them, the absence of such features is not an inconvenience but an exclusion. The pupil with visual processing difficulties cannot use a platform with fixed text and insufficient contrast. The pupil with ADHD cannot sustain attention on a cluttered interface that provides no structured breaks. The implication is that a platform which does not think about inclusion from the outset will inevitably create a two-speed experience: one for typical pupils and one inferior to it for those who most need support. --- ### Data on Pupils' Progress Is Useful Only If It Can Be Read in Three Minutes There is a tendency in educational platforms to offer as much data as possible — detailed reports, progress graphs, engagement indicators, performance comparisons. The temptation is understandable: the data exists, processing it is cheap, displaying it seems like evidence of sophistication. The problem is that a teacher with thirty pupils and five classes a day does not have time to read a two-page report for each pupil. The studies cited by Mabie (2026) and confirmed by the Cognita reports on AI platform implementation in schools globally show that teachers abandon complex reporting features not because they don't value them, but because they don't have time to use them. Data that cannot be read in a few minutes before a lesson is not useful data — it is data that exists. The implication is that a good dashboard is one that answers a single essential question at first glance: who needs additional attention today? The rest — details, histories, comparisons — may exist, but should not be compulsorily visible on first opening. --- ### Training for a Platform Must Begin with Pedagogy, Not with Buttons Carroll, in a 2022 study on primary school teachers' perceptions of their own efficacy during the pandemic, identifies a pattern you recognise immediately if you have ever attended a digital training course: teachers who receive training focused on digital pedagogy — how to maintain a pupil's attention on a screen, how to build authentic discussion in a virtual environment — have a significantly higher rate of platform adoption than those who receive exclusively technical training. You know how the button works. You don't know why to press it. Fang and colleagues (2026) formulate the same idea from the other end: teachers who understand why a tool exists and what pedagogical problem it solves are less anxious about it and use it more creatively. Understanding the purpose precedes mastering the means — not the other way round. The practical implication: a platform that provides access to a tool without also offering a pedagogical framework for how to use that tool in the classroom will be adopted superficially and abandoned quickly. Twelve-minute video tutorials explain the menus. They do not explain why to use them. --- ### The Teacher Does Not Want to Be Replaced — They Want to Be Amplified You will find this theme, formulated in different ways, in almost every study in recent literature about the integration of technology in education. Hanna and colleagues (2026) call it "identity tension" — the conflict a teacher feels when a platform does well something they used to do, and they don't know whether this is good or bad. Celik and colleagues (2026) call it "the perceived threat of substitution" — and show that teachers who perceive an AI tool as a competitor systematically reject it, regardless of its technical qualities. What teachers accept — and what sustained adoption confirms — is amplification. The platform that does what the teacher cannot do at scale: correct a hundred essays in a night, simultaneously track the progress of thirty pupils, be available at eleven o'clock at night when an anxious child wants to do one more test. The platform that does not pretend to read an essay the way a human reads it, to feel what a human feels, to judge what a human judges after eighteen years in the classroom. The implication is that any educational platform that positions itself as a substitute for the teacher will lose in the long run. Any platform that positions itself as the teacher's instrument wins not just adoption, but loyalty. --- *The sources on which this synthesis is based are those from the main article's bibliography. I reproduce them here for direct access:* - Cornerstone Communications & Edsby (2025). *Survey Finds Majority of Schools Using 10 to 15 Educational Apps.* THE Journal. [thejournal.com](https://thejournal.com/Articles/2025/04/01/Survey-Finds-Majority-of-Schools-Using-10-to-15-Educational-Apps.aspx) - Mabie, E. (2026). *10+ Must have top features for every online teaching platform in 2026.* Edsby. [edsby.com](https://www.edsby.com/online-teaching-platform-top-must-have-features/commentary/) - Fang, X., Zhao, J., & Yin, J. (2026). *Primary and Secondary School Teachers' Anxiety in Digital Transformation of Education.* Journal of Teacher Education, 13(1), 95–103. [xbgjxt.swu.edu.cn](http://xbgjxt.swu.edu.cn/en/article/doi/10.13718/j.cnki.jsjy.2026.01.011) - Kortright-Torres, E. E. (2022). *English Language Teachers' Self-Efficacy about Remote Instruction during COVID-19.* ProQuest LLC. [eric.ed.gov](https://eric.ed.gov/?id=ED633650) - Carroll, K. R. (2022). *Primary Teachers' Perceptions of Self-Efficacy during the COVID-19 Pandemic.* ProQuest LLC. [eric.ed.gov](https://eric.ed.gov/?id=ED631843) - Celik, I., Muukkonen, H., & Siklander, S. (2026). *Teacher-AI Interaction: The Role of Trust in Teachers' Acceptance of Educational Chatbots.* Policy Futures in Education, 24(1), 164–188. [fachportal-paedagogik.de](https://www.fachportal-paedagogik.de/literatur/vollanzeige.html?FId=eric_ej1493750) - Unal, Z. (2025). *Enhancing Educator Engagement with AI: The TeacherServer Experience.* SRATE Journal, 34(1). [eric.ed.gov](https://eric.ed.gov/?id=EJ1464416) - Hanna, F., et al. (2026). *Pixels and pedagogy: MISTT's insights into evolving teacher identities when technology comes to play.* Computers & Education, 249. [sciencedirect.com](https://www.sciencedirect.com/science/article/pii/S0360131526000394) - "Ștefan cel Mare" University of Suceava (2025). *The European IDEATE Project.* [usv.ro](https://usv.ro/proiectul-european-ideate-parteneriat-international-pentru-formarea-profesorilor-in-educatie-incluziva-si-digitala/) - Cognita (2026). *AI learning platform rollout.* ChannelLife Australia. [channellife.com.au](https://channellife.com.au/story/cognita-starts-rolling-out-ai-learning-platform-to-schools-around-the-world) --- ## Why I Wrote the Interview-Article the Way I Did: A Conversation with Bogdan About Structure, Style, and Discomfort *This section is not part of the main interview. It is something different — a conversation about the conversation, if you like. About why we chose to write the article the way we wrote it, and not otherwise. I include it because I think a reader who understands why a text is constructed the way it is has more freedom in relation to it than one who receives the text as a fait accompli.* --- Bogdan Ignat rang me the day after I had published the first version of the article. Not to complain. Not to thank me. He rang to ask me a question I hadn't anticipated: why did you write about your son? What does he have to do with our platform? I sat for a few seconds with the phone to my ear, listening to the brief silence after the question, and I realised that my answer to that question is, in fact, the entire article. That if you don't understand why my son is there, asleep with the grammar book on his chest, you don't understand why anything else is there either. I suggested we continue the conversation in writing. He agreed. What follows is that conversation, tidied of digressions, but unaltered in substance. --- ### "Why did you write about your son? What does he have to do with our platform?" I sat with this question longer than I perhaps should have. Not because I didn't know the answer, but because my answer seemed, put simply, almost naïve: because you built this platform for him. Not for him specifically — you didn't know he existed. But for that generic thirteen-year-old sitting at a desk at night not knowing whether they are ready. My son is that child. And if I don't put him in the text, the article becomes a description of a product. If I do put him in, it becomes a description of a real problem that the product is trying to solve. There is a difference between these two things, and it is greater than it first appears. **Bogdan:** I understand the logic. But there is a risk there — the reader who doesn't have children of that age may feel you are speaking to someone else. Yes. The risk exists. And I accepted it. This choice — of anchoring an advertising text in the personal experience of a father — inevitably excludes readers who do not find themselves in that experience. But it includes something else: authenticity. A text in which you recognise that the author has a real stake, not a rhetorical one, in the subject is more credible than a neutral text that could have been written by anyone. Complete neutrality in advertising does not exist — there is advertising that pretends to be neutral and advertising that declares where it is speaking from. I chose the second. **Bogdan:** And did you consider that perhaps the reader feels you are too much in the text? That the article is more about you than about the platform? I did consider it. And I checked, paragraph by paragraph: there are sections of the article in which I disappear entirely and speak exclusively about the platform — how the assessment works, what it can and cannot do, what the documented limitations are. There are sections in which I return as an observer. And there are the opening and closing moments in which I am present as a person, not as an advertising writer. This structure — personal entry, documented body, personal exit — is deliberate. It is the way in which a reader with no prior connection to the subject can enter through the door of human experience and leave with verifiable information. Had I reversed the order — technical information at the entry, personal note at the end — the right reader for this article would have given up after the first paragraph. --- ### "Why did you include the studies showing that AI grades poorly? Isn't that counterproductive for us?" This is the question I was truly waiting for. Not because I feared it — but because it is the right question, and I was glad he was the one asking it, not me. **Bogdan:** Seriously, from a purely commercial perspective, why did you include the studies showing that automated assessment has moderate to weak correlations with human assessors? Doesn't that put readers off? Do you know why I included those studies? Because I didn't have to ask your permission to include them. You mentioned them yourself, on your own initiative, during the interview. Someone who knows the limitations of their product and names them before being asked is either an extremely honest person, or one who knows that those limitations will come to light regardless and prefers to control the narrative himself. Or both, which is more likely. And in the text, when I include those studies, something specific happens: the reader who knows they exist and goes looking for them finds them already in the article. They no longer have reason to believe I concealed something. And the reader who didn't know those studies existed receives a more complete picture of reality, which allows them to make a more informed decision. Both scenarios are better for the platform's long-term credibility than an article that omits them. **Bogdan:** But there is a threshold. If I include too many limitations, people stop buying. Yes. There is a threshold. And I judged that threshold not to be exceeded in the article. The limitations I included are real, academically documented limitations, with verifiable sources — not limitations invented to appear balanced. And in every case where I mentioned a limitation, I immediately explained why the platform's architecture reduces that specific risk. I didn't leave a limitation hanging in mid-air. I said: the risk X exists, and the platform has addressed it through mechanism Y. That doesn't put off an informed user — it convinces them that the product was designed by people who understand the problem. **Bogdan:** And if mechanism Y doesn't work well enough? Then the article will be contradicted by users' experience, and I shall no longer be able to write credible articles about the platform. That is the implicit contract of an honest advertorial: if the product disappoints, the author pays with their credibility. That is why it mattered to me that the limitations I included be real — because if I overstate the platform's performance, the consequences are mine, not yours. --- ### "Why did you choose this style? An advertorial can be written much more simply" It can. And a simpler version would probably be more effective on certain metrics — reading time, retention rate at technical paragraphs, number of clicks on the platform link. I considered the simple version. It looked like every other text about educational platforms: three paragraphs of benefits, a bullet point of features, a call to action. **Bogdan:** And why didn't you choose it? Because cojocarupetru.info is not a marketing site. It is a site with a specific editorial voice, which readers who arrive there recognise. If I write an advertorial in that voice that sounds like every other advertorial, I create a dissonance the reader senses immediately, even if they cannot name it. They sense that the author has switched to a different register, that they speak differently about something they are selling compared with how they speak about everything else. That dissonance is toxic for credibility — more toxic than the style itself. **Bogdan:** That style — with the rain, the sleeping son, the morning coffee — isn't it itself a form of manipulation? Don't you use narrative to create an emotional state favourable to the product? Yes. And no. Yes, in the sense that every text, through its structure, its tone, through what it chooses to include and what it chooses to leave out, produces a state. There is no emotionally neutral text — even the driest technical manual is not neutral; it produces a state of numbness and disinterest, which is itself a state. The question is not whether a text produces an emotional state, but whether the emotional state it produces is honest in relation to its content. The state my article produces is one of reasonable concern about a real problem — the preparation of a child for an important exam — and of moderate hope about a partial solution. Not euphoria. Not certainty. It is closer to what you feel when you find something useful, not something miraculous. That is the appropriate state for your product, which is useful, not miraculous. Had I produced a state of euphoria or absolute certainty, I would have manipulated in the bad sense of the word — I would have produced a state inappropriate to reality. I don't think I did that. **Bogdan:** Do you realise you give very long answers to short questions? I do realise. It's an occupational habit. Or a style. I don't know exactly where one ends and the other begins. --- ### "Where did the section about the format training risk — the second morning — come from?" It came from the discomfort of having published something incomplete. That is the most honest answer I can give. I had published the first version of the article and felt good about it — it was honest, it was documented, it named the limitations. And yet the next morning, with my coffee in front of me and my son sleeping in the next room, I realised something was missing. Not a marginal piece of information. The most uncomfortable question possible was missing: what happens if your platform makes children better at tests and worse at writing? **Bogdan:** And why did you include it? The article was already published. Because I couldn't leave it out once I had thought of it. Once I had conceived it, omitting it would have been an active choice — not a passive omission, but a decision not to say something I knew to be relevant. And if I make that decision, the contract with the reader is broken. Not immediately. Not visibly. But I know I have broken it. **Bogdan:** That risk exists with other preparation tools too — tutoring, exercise books, any repetitive format. Why did you treat it as a risk specific to the platform? I didn't treat it as a risk specific to the platform. I treated it as a risk specific to any preparation focused exclusively on a fixed format — and I stated explicitly that this applies to any training tool, not just ltrace.ro. But I included it in the article about ltrace.ro because the article is about ltrace.ro. A reader who pays for tutoring with a human teacher will judge that risk themselves — they have a different type of relationship with the information. A reader who pays for access to a digital platform may end up using it exclusively, precisely because it is available at any time and requires no travel time. The risk of exclusive use is greater for a digital platform than for traditional tutoring, and that is why it merited mentioning in context. --- ### "Is there anything in the article you disagree with?" This question was mine, not his. At the end of the conversation, after I had answered all his questions, it seemed right to reverse the direction. He paused for a moment. This man — who had built a platform over years, who had spoken to me for hours about architectures and databases and academic limitations — paused before answering. **Bogdan:** The sentence "I have never seen a ten-year-old willingly do the same test three times with a traditional exercise book." You use it as an argument in favour of the platform, but immediately afterwards you add that increased motivation towards the wrong goal can be more dangerous than a lack of motivation. That leaves the original sentence in an ambiguous position — the reader doesn't know whether to read it as praise or as a warning. He is right. The sentence is ambiguous. And the ambiguity is intentional — not as a rhetorical technique of confusion, but because the reality is ambiguous. That a child does a test three times is simultaneously a good sign and a potentially bad sign, depending on why they are doing it. If they do it to understand where they went wrong, it is good. If they do it to memorise the answer, it is bad. The platform cannot control with what intention the child comes to it. And nor can I, writing the article, know with what intention the reader's child comes to the platform.[![The Teacher Who Never Sleeps](https://cojocarupetru.info/images/Advertoriale-sound/logo-hint-one.png "The Teacher Who Never Sleeps")](https://hintstudio.ro/) Perhaps I should have said that more explicitly in the text. Perhaps that ambiguity is a failure of clarity rather than a successful editorial choice. I believe Bogdan when he says the sentence sits oddly there. It is the best observation anyone has made to me about the article. --- *The conversation above took place through written messages, over the course of two days. It has been edited for coherence and length, but not for content. Both voices — mine and Bogdan Ignat's — are reproduced with his explicit consent.* ---