The Accidental Teacher
I Build Around the Child
Meet the homeschool dad behind the tools.
I'm Scott Somers, the Accidental Teacher. I'm a homeschool dad, a former middle-school teacher and a curious builder who keeps coming back to one question: how do we make an education that takes this particular child seriously?
I want the answer to work in a real home. A parent needs something they can use, and a child needs a reason to engage with the work in front of them. Getting there takes attention, effort and a willingness to change what isn't helping.
That is what drives me to build courses and learning tools, learn new skills and use AI. I care about what those tools make possible, and I care about the work required to use them well.
When I share something I've made, I want you to know the person behind it: a dad working through education with his own family, trying to make something useful for yours.
I'm in the trenches, too
I live in Glennallen, Alaska, by choice. My wife and I have roughly fifteen years of homeschooling behind us, through different combinations of homeschooling, school and finding our way toward what fits. I also spent two and a half years teaching middle school in the public-school system.
I know there are caring, hardworking adults on both sides of a school door. I also know that loving a child does not automatically tell you what to do about tomorrow's lesson.
When a parent says a curriculum looked wonderful until they actually tried to use it, I understand the frustration. When they wonder whether to buy another program, change their approach or give the current one more time, I take that question seriously. The decision has to work in their home, with their child.
I am making those kinds of decisions with my own family while I build things intended to help other families. There are instructions to improve, assumptions to question and choices that take more than a confident answer to resolve.
I want that to come through when people meet me online. I am right there in the trenches with other homeschool parents. Their frustrations deserve to be understood by somebody trying to help.
Start with the person
Before I want to know which program a family is considering, I want to know the child.
What does this child hope for? What do they dream about? What frightens them? What do they love, and what can they hardly stand? What do they want to understand, make or become?
Those questions give us somewhere useful to begin. A list of subjects and grade levels can tell us what needs teaching. It leaves a great deal to discover about the person who will be doing the learning.
My son wants to become a chef. His chemistry course is built around cooking. Chemistry has a place in something that already matters to him. The course still needs to teach substantial chemistry, and we still have to do the teaching, thinking and checking. His interest gives us a way into the work.
Same standards, different way in.
I recognize the importance of that connection from my own life. I hated math until science gave it a reason to exist. Later, I taught middle-school math. A subject can look very different when you have a reason to care what it helps you understand.
Children still need practice. They need to meet ideas beyond their existing interests and persist when something is difficult. An interest can open a door into demanding work, including work they would never have chosen at first. Individualized education keeps paying attention to the person doing it.
Being a beginner helps me remember
I like building things. I like the point where an idea has to survive contact with something real.
That curiosity takes me into coding, curriculum, writing and learning to play a cigar box guitar. Being a beginner is useful company for an educator. An instruction can make perfect sense to the person giving it and leave the person receiving it wondering where to put their hands.
I want to remember that feeling when I write instructions for somebody else's child.
My Christian faith is underneath this work as well. It shapes how I understand a person's worth and my responsibility toward them. A child's dignity is present before they finish an assignment or demonstrate a useful skill. Helping them learn calls for care, honesty and expectations that take their capabilities seriously.
That gives me a reason to be patient with a person and rigorous about the work I put in front of them. I am interested in what a child might become. I also want to know the child who is here now.
Making it doable
Individualized education can sound wonderful and still leave a parent wondering where to begin. Nobody needs another educational ideal that turns into an impossible assignment for the adult.
I would start small: one child, one learning goal and one interest that already matters to them. Identify what the subject actually requires. Then look for a task that connects that requirement with something the child cares about. You can adapt a useful resource you already have. Every family does not need to build an entire curriculum from scratch.
Try the task. Ask your child to explain what they did and why. Watch where they can reason through a problem and where they are following directions without understanding. Use what you observe to decide what to explain, practice or change next.
That cycle gives the adult something concrete to work with. It also leaves room for a child to surprise us. An approach that looked excellent in the plan may need revising once the person it was designed for gets involved.
AI can help me develop that plan. I can use it to explore approaches, draft a sequence, create variations and identify gaps to investigate. Work that once would have been difficult to attempt becomes more practical to develop and revise.
Then comes the judgment. Does the sequence make sense? What knowledge does it assume? Is the explanation accurate? Does the activity teach the intended idea? What would show that the child understands?
Those questions take time to answer well.
The work after the first answer
When people hear “AI user,” some immediately picture sloppy, error-ridden or lazy work. They see the generated page and imagine that getting it onto the screen was the whole job.
That misses a large part of what I spend my time doing.
I have to decide what I am trying to build and describe it clearly enough to begin. I have to check the sources behind claims, work through tasks, inspect the result and revise what fails. Sometimes a clever suggestion needs to go because it does nothing useful for the child. Sometimes the instructions need another pass because the person writing them knew more than the person reading them.
A polished page can still contain a false fact, a weak explanation or an activity with no useful purpose. Making it look finished is much easier than establishing that it is ready to use.
I understand why careless AI output makes people skeptical. Sloppy work deserves criticism. I want the care, time and responsibility involved in doing this well to be visible, too. Using a powerful tool leaves plenty of work for the person holding it.
A whole fleet of interns, all billed together
People often compare AI to an intern. My experience feels more like managing a whole fleet of interns, all billed together.
I can divide the work among different helpers. Explore the subject. Propose a structure. Examine an explanation. Look for an assumption I missed. Challenge the result.
That makes more work possible, and it creates another responsibility: making sure the pieces belong together and deserve to be used.
Five helpers repeating the same mistake would give me a very efficient mistake.
Agreement among AI systems does not make a claim true. They may share assumptions, training influences or a mistaken source. A second answer becomes useful evidence when I can trace it back to an original source and check what that source actually says. A plausible citation on a generated page needs checking as much as the sentence beside it.
NIST's guidance on generative AI describes risks including confidently false output, harmful bias, overreliance and correlated failures. I take those warnings as reasons to build checking into my work with multiple helpers. Adding more helpers does not establish reliability. NIST, Generative AI Profile
The tools have costs and limits. My capacity to supervise them has limits, too. Choosing work I can actually evaluate is part of using them responsibly.
I remain responsible for what I publish, teach or hand to a family. Delegating part of the process does not move that responsibility away from me.
We need to know what we're using
A blanket declaration that all AI is evil gives families very little help understanding a technology they are already encountering. Uncritical excitement is equally poor preparation.
Parents who ask hard questions are doing something valuable. I want us to know enough to make informed choices about where a tool belongs in our children's learning and where it should be left out.
That includes understanding its biases. An AI system can treat one cultural perspective, family situation, vocabulary or set of abilities as the default. It can offer advice that sounds reasonable while assuming circumstances that do not fit the household receiving it.
My questions can carry assumptions, too. If I describe a child entirely through what is going wrong, I have narrowed the picture before the system generates a word.
I need to examine the question as well as the answer. Whose experience is missing? What has been assumed? Where did this claim come from? What would make me reconsider it? What does the actual child show me that the tool cannot know?
UNESCO's education guidance puts human agency and accountability at the center of using generative AI. It also calls for evaluating possible uses for their purpose and the people involved. That gives us a useful question to keep asking: how does this particular use serve this particular learning need? UNESCO, Guidance for Generative AI in Education and Research
The child still has to do the thinking
I want to be careful about the difference between helping a child learn and helping a child produce an answer.
In one mathematics experiment at a Turkish high school, access to a general AI interface improved assisted practice results, but that group performed worse than the control group when the assistance was removed. A teacher-informed tutor designed to offer help through hints largely avoided that decline. It did not establish a clear independent-test advantage. The design of the help mattered. Bastani and colleagues, PNAS
There is promising evidence as well. A study of two purpose-built lessons for university students in introductory physics at Harvard found stronger immediate learning gains with an AI tutor than with the comparison classroom lessons. That was a particular tutor, two lessons and a particular group of people. It gives us something useful to investigate, rather than a universal conclusion about children or classrooms. Kestin and colleagues, Scientific Reports
I want to keep that distinction in view when I build and when I parent. A tool might help me prepare a better explanation or help a child find a starting point. The child still needs opportunities to reason, make choices, explain an answer and discover where their understanding is incomplete.
Their writing needs to remain their writing. Their decisions need room to be their decisions. I want a finished assignment to tell us something honest about what they can do.
That takes judgment from adults who know the child, along with a willingness to change our approach when the evidence is inconvenient.
Why I'm here
I want more families to have practical ways to build an education around the child they know and love. That takes listening to parents, taking their frustrations seriously and making things they can actually use.
It also takes adults who are willing to keep learning. I want to work with people who care enough to ask hard questions, try thoughtful ideas and examine what happens. That is the kind of work I want to contribute to, and the kind of care I want families to recognize in what I share.
I bring my experience as a dad and a teacher, my faith, my curiosity and a willingness to revise. I want to understand what makes a child tick and find a way to connect that person with substantial learning.
AI expands some of what I can attempt. The child gives the work its purpose.
So when you meet me through an article, a learning tool or a video, I hope you can see the person doing the work. I am Scott Somers, the Accidental Teacher, and a homeschool dad who wants to make learning fit the people we are raising.
What is something about your child that you wish their education understood better?