What Should School Do With the Time AI Might Save?

1. Guest Background

Joe Liemandt approaches education as a principal with a product and entrepreneurship background. In this roughly 56-minute episode of a16z / The Ben & Marc Show, whose metadata names Marc Andreessen and Ben Horowitz as hosts, he describes Alpha School through the experience it promises students and the organizational work needed to deliver it. He says his early priorities as principal included making the model understandable, packaging it as a coherent offering, and finding a way to expand without sacrificing quality.

His personal account helps explain that emphasis. Liemandt recalls skipping school, doing the minimum required to enter Stanford, and later dropping out to start a company. Once building something he cared about, he says, he worked approximately 100 hours a week. He also remembers Stanford as valuable for encounters with technology entrepreneurs and for building with friends who became part of his company. Those experiences inform his preference for agency, relationships, and consequential work.

They remain personal experiences rather than evidence that dropping out or working extreme hours improves education. The distinction matters throughout the episode: Liemandt is explaining and advocating a model he operates. Other participants also discuss an academy and their ambitions for university education. Their proposals should be read as separate contributions, rather than attributed indiscriminately to Alpha or treated as independently evaluated outcomes.

2. What the Episode Covers

Alpha’s organizing question is what students should do with a full school day if academic instruction can become substantially more efficient. Liemandt describes three commitments: students should love school, complete their academic work in two hours a day, and spend the remaining time developing life skills. The afternoon therefore has a defined educational purpose. Adults lead peer projects involving teamwork, leadership, entrepreneurship, financial literacy, relationships, communication, and persistence. The proposal retains the school as an all-day institution while reallocating its time.

Two different efficiency claims sit behind that proposal. Liemandt says cognitive science, learning science, and large language models make an engine capable of teaching ten times faster than conventional classroom instruction. His description of Alpha’s commitment is different: students learn twice as much in two daily hours as they would in six classroom hours plus homework. One is a claim about technical potential; the other is a claim about the school’s offering. The conversation does not supply the comparative measurement needed to establish either independently, and the two should not be combined into a single demonstrated result.

The university discussion asks what academic preparation is supposed to enable. Liemandt reports that more than 70 summer interns from MIT, Stanford, and UT expressed widespread concern about whether university was preparing them for life after graduation. He recounts their fear that AI would change the entry-level jobs they had expected. Their response, in his telling, was an intensive project: researching learning science, working with a data structure called Brainlift, and building an educational language model they hoped could serve people globally for free. He says they worked 80-hour weeks. The educational point is the connection between study and construction: research became necessary to making something useful.

The participants also discuss deliberately chosen challenges outside ordinary coursework. Liemandt describes World School’s year of service in Africa and an agreement acknowledging its difficulty. An academy participant separately says students have a $5,000 travel budget that they decide how to use. His hope that many travel alone to countries with GDP per capita below $20,000 is a personal aspiration, not a stated universal requirement. These examples express a commitment to independence and unfamiliar environments; the episode does not evaluate their outcomes systematically.

Entrepreneurship provides the most explicit commercial version of the model. Liemandt says Alpha begins financial literacy in kindergarten, using a school currency, alphas, to practice earning, saving, spending, investing, and donating. Founder School extends that progression into business building. He describes a guarantee that students will start a business earning $1 million in profit during high school or receive their tuition back. He also outlines milestones rising from $100 through successive tenfold steps to $1 million. The school, he says, opened in August, and he predicts some students will reach the largest target in their first year. That prediction is not a reported fulfillment of the guarantee, whose contractual and accounting conditions are not supplied.

3. Core Views: Reasoning, Examples, and Limits

The strongest organizing idea is that educational technology needs an institutional design around it. Liemandt predicts that inserting Alpha’s learning engine into an otherwise conventional schedule would fail. His reasoning links several elements: academic work must feel achievable, saved time must become worthwhile activity, and adults must organize challenges students want to undertake with peers. His account of happiness follows the same logic. High standards can be attractive when students experience themselves becoming more capable together. The episode offers a plausible mechanism to investigate, rather than a causal demonstration that this combination produces the reported satisfaction.

The title’s comparison with vacation comes from Liemandt’s account of an internal survey. He says more than 90% of students love school, half would prefer school to vacation, and one campus reaches 65% on that comparison. These figures reveal what the school values and what its principal reports. Without sample size, response rate, question wording, or an account of who was included, they cannot establish how representative those preferences are. They also do not isolate which part of the school experience students value. Enthusiasm for friends, projects, academic progress, or the institution as a whole could require different explanations.

The academic mechanism is more specific than the headline. Liemandt describes assigning work by mastery rather than age: an eighth grader missing fifth-grade knowledge receives the earlier lesson. The engine aims for an 80–85% correct-answer range, adding scaffolding when work becomes difficult and accelerating when it becomes easy. The rationale is that unresolved gaps can make a time-based curriculum progressively more discouraging, whereas attainable challenges restore opportunities to succeed. But an immediate accuracy target describes task selection. It does not establish retention, conceptual understanding, or the ability to apply knowledge independently in an unfamiliar setting.

The escape-room example makes the life-skills proposal similarly tangible. Liemandt says third graders wear microphones, AI analyzes their conversation, and initially less than 20% of their language supports teammates. After an eight-week workshop, he says, passing requires 85% team behavior, while completion becomes three times faster. The useful design principle is a feedback loop: observe behavior, practice deliberately, and assess again. The measurement boundary is equally important. Encouraging language is only part of cooperation, and faster completion need not result entirely from improved teamwork. The episode does not explain how the language measure becomes a team-behavior score or how improvement transfers to different tasks.

Liemandt also says fourth and fifth graders must pass a Wharton Everest teamwork simulation that approximately half of MBA students fail. This illustrates his willingness to set demanding expectations for children. It does not establish that the children have better general teamwork skills than MBA students: the conversation provides no common testing conditions, versions, or scoring comparison. The defensible takeaway concerns curriculum design—life skills have a passing requirement—rather than a ranking of populations.

The university discussion distinguishes components of educational value. Liemandt recounts that Alpha High’s first graduates all had 4.0 GPAs early in college, yet criticized introductory lectures; five wrote newspaper editorials. They preferred seminars, clubs, and activities. Their experience supports examining the university as a bundle of experiences, but the cohort size and course differences are unspecified. An academy participant adds a crucial qualification: K–8, high school, and university require different designs. Agency and meaningful relationships may be shared goals without implying identical schedules or responsibilities.

Liemandt’s term cognitive onload describes another proposed AI benefit: acquiring the knowledge needed to build. He claims tenfold acceleration and the possibility of gaining master’s-level knowledge in a year, reports high-school students rebuilding the Timeback software stack, and predicts that students will write most learning software within two years. A concrete example is Austin’s AP curriculum, combining historical geography guessing, tier rankings, and a snarky AI. Liemandt says all participating students earned AP scores of five, but gives neither their number nor a comparison group. The example shows an approach to engagement; it cannot by itself establish why the scores occurred or validate the wider forecast.

Finally, motivation and scale require careful attribution. Liemandt’s description of schooling as 90% motivation is an educational judgment, not a measured allocation of causal influence. His explanation of parents’ reports of rapid transformation is similarly interpretive. He also says campuses grew from 12 to 50 while scores, satisfaction, happiness, and workshop quality improved over four years. Yet his recruitment strategy explicitly selects families aligned with the model and prepared for early problems. That alignment may help a school function while complicating evaluation through self-selection. Without comparable measures and student populations, improvement during expansion does not establish improvement caused by expansion. The academy side’s reported employer concerns—people skills, maturity, and systems thinking—offer further design priorities, but no sampling information makes them a universal diagnosis of graduates.

4. Learning and Application

For a school leader, the practical starting point is a bounded redesign with two linked questions: how will students learn the academic material, and what will they do with any time released? A pilot could track starting knowledge, scaffolding, completion time, and later recall in one unit while introducing a purposeful peer task. Keep Liemandt’s tenfold potential claim separate from Alpha’s two-hour commitment. The 80–85% accuracy range can be examined as part of his design, but should not become a universal target without checking independent explanations, reliance on hints, and performance on unfamiliar problems.

Life-skills work can borrow the escape-room feedback structure without assuming its reported numbers will recur. Define observable behaviors before the activity, gather examples, coach specific weaknesses, and reassess. Include actions as well as language: whether students invite contributions, integrate disagreements, and follow through on responsibilities. Otherwise, a scoring system may reward supportive phrases without improving coordination. A different task or new group provides a useful transfer check. Adults still need to interpret what happened; the episode does not establish that AI scoring alone can judge the full quality of collaboration.

For university programs, the intern project suggests organizing study around a real intended use. Specify what students must understand to build the product, which relationships they must manage, and what evidence they must present at reviews. Require both a working result and an explanation of the reasoning behind it. This gives theory a purpose while preventing delivery pressure from making understanding optional. Handling rejection, reorganizing work, and responding to feedback can also enter assessment. Liemandt’s reported 80-hour weeks should remain a feature of that account, not a required measure of commitment. Challenge and support need redesigning for each age group.

Families evaluating a new school can ask to see an ordinary day, examples of support when students struggle, and the basis for reported outcomes. Liemandt’s admission that he took two years to move his own children, alongside his account of New York growing from 10–12 students to approximately 150, illustrates the distance between interest and adoption. Visits and conversations with enrolled families may reduce uncertainty, but they do not remove selection effects. A school’s emphasis on philosophical fit should help clarify expectations about AI use and experimentation while leaving room to question results.

Entrepreneurship programs warrant equally concrete questions. Start with affordable tasks that connect financial records, customer responses, and subsequent decisions. When assessing Founder School’s offer, distinguish revenue, cash received, milestone earnings, and profit. Ask which costs count, what period the guarantee covers, and what triggers a refund. Those details are absent from the episode; forecasts about rapid progress cannot substitute for verified fulfillment. A failed small venture can still produce learning, but that is a different judgment from whether a particular tuition contract offers value.

One networking lesson is immediately usable. Liemandt recounts a 14-year-old reaching Jensen, receiving 90 seconds, and then being criticized for asking questions that preliminary research could have answered. Before an interview, students can assemble public information, identify unresolved questions, and explain why this particular person’s perspective is needed. Afterward, they can turn criticism into a revised preparation process. Access may raise ambition, but research, question selection, and response to feedback are the more transferable skills. A reply from a prominent person should not become the required definition of success.

Source

• Original episode: The AI School Kids Love More Than Vacation | Alpha School's Joe Liemandt