The first riddle was not artificial intelligence. It was arithmetic. Caleb Hicks had more than 300 students moving through his keyboarding classroom at Lehi Junior High School, sometimes seven or eight periods in a day, with 42 desks waiting in each period. There were only so many names, moods, questions and half-finished thoughts that one teacher could hold in his head. He learned the most engaged students. He learned the hardest students. In between sat a large and largely anonymous country.
Hicks has a name for it now: the middle 80 percent. The phrase is imperfect mathematics, but exact memory. These were not average children. They were simply the ones whose needs did not arrive with a raised hand, a perfect score or a disruption loud enough to interrupt class. Hicks has said he thinks about them because he once belonged there himself.
The entire arc of his career can be read as a series of attempts to improve that ratio. He turned typing drills into games. He built curriculum for people learning to code. He joined Apple, co-founded an online school and started a career program for teenagers. Then, in early 2023, he co-founded SchoolAI in a room above his garage in Lehi. Generative AI had supplied a new instrument. The old problem remained in charge.
The figure is Hicks's own shorthand for the attention gap he experienced as a teacher.
A classroom with permission to misbehave
Keyboarding is not an obvious birthplace for an education company. It is repetitive by design, a subject built from home rows and dutiful fingers. Hicks refused to leave it there. He began classes with free writing on blank keyboards. He challenged students to beat his typing speed and rewarded the winners with time to play computer games. The bargain was shamelessly concrete: become faster than the teacher, earn a small liberty.
Soon the lab widened. Hicks introduced career exploration and teen entrepreneurship. Students started businesses and worked toward earning $1,000. Their progress in class reflected progress in the enterprise. He later described the classroom as a lab, which catches both his instinct and his appetite for risk. He was willing to make an assignment slightly unruly if the alternative was tidy indifference.
“My classroom was a lab.”Caleb Hicks
There was biography in the choice. Education was the family business: Hicks has said his parents and siblings were teachers. At Weber State University, where he studied business and marketing education, he prepared to join them. In Lehi he also helped charter Utah's first middle-school FBLA chapter. The local program grew beyond 100 students and sent nine of them to a national competition. Commerce, technology and teaching were already sharing a desk.
The common element was consequence. A student could type fast enough to win something, build a real business or discover a career before paying tuition to pursue it. Learning, in Hicks's preferred version, was not a rehearsal for a distant life. It had to touch the present.
The curriculum gets a passport
Hicks left the junior-high classroom without leaving curriculum. At DevMountain, the Utah coding bootcamp, he led an immersive iOS program and built project-based instruction around tools including Swift, Firebase and Core Data. At Apple he became a senior manager in worldwide education marketing, leading instructional design for K-12 schools and universities. The scale changed. The question did not: how could one lesson bend toward different interests without breaking?
He gives a characteristically commercial example. Supply and demand might be taught through a sneaker drop for one student and a comic-book shop for another. The economic principle stays still while the route into it moves. Personalization, in this telling, is less about making the work easier than giving curiosity a handle.
In 2017, Hicks co-founded Lambda School, later renamed Bloom Institute of Technology. As president, he oversaw the academic operation from admissions and onboarding through learning design, delivery and student services. The school became known for income-share agreements under which students paid after landing jobs above a set salary. It was ambitious, contested and operationally revealing. An online school could see a great deal about a learner's progress, but only if its systems were built to notice.
After Lambda, Hicks founded Factor in 2021. The program let teenagers test-drive careers through hands-on projects with working professionals. He later spun it out as a nonprofit. Once again, the premise was impatient with abstraction: try the future before committing to it.
The spelling list and the sidekick
When ChatGPT appeared, Hicks tried it on an ordinary domestic problem: his daughter had spelling words to learn. He asked the system to make stories using those words in settings she enjoyed. The assignment stopped feeling generic, and she did well on the test. The experiment was small enough to fit at the kitchen table and large enough to expose the possibility he had been chasing for years. A lesson could adapt quickly, not merely to a grade level, but to a person.
SchoolAI's first work for a district was less whimsical. Schools wanted the capability of ChatGPT without casually handing student identities to a public chatbot. Hicks's team built a privacy layer to shield personally identifiable information. Teachers remained in control of the activity and its boundaries. The point was not to turn a classroom loose inside a general-purpose model. It was to make the model behave inside a classroom.
During testing, one teacher used the system to create an interactive text adventure about the Battle of Yorktown. Students did not merely read prompts. They sent hundreds of messages while the teacher watched their historical understanding take shape. The unexpected experiment helped produce what SchoolAI calls Spaces: teacher-designed learning environments in which an AI sidekick can role-play, question, coach and adapt while a dashboard tells the educator who is moving and who is stuck.
“AI is not the thing, it's the thing that gets you to the thing.”Caleb Hicks
That aphorism is Hicks at his most useful. It prevents the software from becoming the story's hero. Consider a student from Afghanistan whom Hicks met at a school where 24 languages were spoken. The student spoke Dari. He had arrived eager to ask questions, then became aware that repeated requests placed a strain on busy adults. A Dari-speaking AI sidekick offered patience without social cost. It helped with classwork, then helped him find words for conversations about soccer at lunch. The result that mattered happened between people.
A venture-sized answer to a human-sized problem
SchoolAI entered schools during an awkward season. Many districts were banning generative AI just as teachers and students were beginning to use it. Hicks did not dismiss their caution. He argued for intentionality: the technology should push a student to learn, not furnish a shortcut around learning. Teacher control, moderation and privacy were not garnish. They were the admission price.
Adoption came quickly. By April 2025, the company said its tools were used in more than one million classrooms across all 50 states and more than 80 countries, supported by strategic partnerships with over 400 school districts. A $25 million Series A brought its reported total funding to $32 million. The company planned to deepen its long-running AI sidekicks, improve the learning data available to schools and expand beyond K-12.
Those figures make a handsome fundraising slide. They also conceal the company’s hardest obligation. A dashboard that claims to show who needs help can influence a teacher's attention. A chatbot used by children must know its boundaries. A personalized assistant that follows a learner over time acquires both usefulness and gravity. Hicks speaks enthusiastically about infrastructure because the work is infrastructural: privacy, moderation, data and judgment, all humming beneath the visible conversation.
He insists that the teacher becomes more important as AI enters the room. That is partly philosophy and partly product strategy. SchoolAI is built to return information to an educator, not to hold it in a private exchange with a machine. The software may translate material, vary the reading level or provide another example. The teacher decides what happens next.
The view from the back row
Hicks is now a founder, chief executive, occasional angel investor and newsletter writer. Yet his public language remains crowded with classroom verbs: see, hear, reach, notice. Even the ambition he gives to parents is strikingly modest. He wants them to ask, “How was school today?” and hear that school was awesome, followed by a reason.
It is tempting to measure his story by its loudest numbers: the funding round, the countries, the classroom count. The quieter measure is more revealing. Can a teacher discover, before the test and before the semester ends, that the student in the middle row does not understand? Can the lesson arrive through an interest the student already owns? Can software give an adult more time to behave like an adult?
The keyboarding lab did not give Hicks an answer. It gave him a durable irritation. Hundreds of students could pass through one room while most remained only partly known. SchoolAI is his most elaborate response so far, built with language models, venture capital and the operational scars of several educational experiments. Beneath all that machinery sits an old teacher's wish: fewer anonymous children before the bell.