Make School: income share agreements and the future of higher education

Executive overview

Most colleges charge tuition upfront regardless of whether graduates find work, creating a misaligned incentive. Make School flips this: the school takes on debt when students enroll and is only repaid if students get jobs.

The result is a fully project-based, accredited CS bachelor's degree — completable in two years — built around career outcomes for low-to-mid-income students who cannot afford to treat college as a luxury.

The core insight: when a school only gets paid if its students succeed, every curriculum and hiring decision becomes a bet on that outcome.

The Make School model

• Students pay nothing upfront; Make School borrows to cover the cost of education
• Tuition is repaid only after graduation and only if the student is employed
• Bachelor's completable in two years via a structured 40-hour work-week schedule
• All classes are project-based — no traditional problem sets; learning is output-driven
• Faculty are industry engineers acting as live mentors and unblockers, not lecturers
• Curriculum is accredited by the same body (WASC) that accredits Stanford and UC Berkeley

Why project-based learning matters for outcomes

• Portfolios built from week one give recruiters concrete evidence of skill, not just credentials
• Teaching the same concepts across multiple languages and frameworks lets students distinguish language-specific quirks from universal CS theory
• Students routinely land jobs in stacks they did not study — iOS developers hired as Android engineers, web devs hired in different languages
• The "flipped classroom" model means curriculum is pre-written; class time is spent unblocking students in real time
• A 30-minute session with a mentor can replace hours of solo debugging — Jeremy estimates a 10x speed multiplier

The autodidact myth and survivorship bias

• MOOC completion rates of 2–5% show that ~95% of learners are not self-directed by default
• The tech industry over-represents autodidacts because structured education has historically been poor, not because autodidacts are more capable
• The best programmers visible in industry mostly come from privileged backgrounds with safety nets — selecting for survivorship, not raw ability
• Most people who could become excellent engineers need mentorship and structure to get there
• The perception that someone "just can't code" often reflects bad curriculum and poor teaching, not lack of ability

Elite education as a luxury good

• 60–75% of students at Stanford and MIT come from the top 20% of families by income; up to a quarter come from the top 1%
• Liberal arts education without career focus was designed for and by elites — it works when opportunities are guaranteed by network and background
• Make School's student body is a near-demographic inverse of Stanford: majority low-to-mid income
• Telling low-income students that degrees are optional is true mainly for those with privilege and safety nets
• The "drop out and succeed" narrative (Gates, Zuckerberg, YC founders) reflects the same survivorship bias — Jeremy's own YC batch had zero surviving companies founded by young dropouts other than Make School

How ISAs change institutional incentives

• ISAs force a school to think daily about student outcomes rather than campus renovations, sports, or research prestige
• Predatory ISA misuse: replacing existing scholarships with ISAs increases total student cost without aligning incentives
• The most powerful ISA model is full-commitment — taking no money upfront so the school is entirely dependent on graduate employment
• 90% of Make School students choose full ISA; Jeremy's target is 99%
• ISA is applicable far beyond CS: nursing, teaching, and other fields with unmet employer demand are natural candidates
• A college willing to offer ISA for some programs but not others reveals its own confidence in what those programs actually deliver
• Core principle: students should not pay if they are not successful — whether via ISA, loan forgiveness, or another mechanism

Degrees, employer reality, and the equity gap

• PR from Google and Apple about dropping degree requirements does not reflect recruiting practice on the ground
• University recruiters have assigned school territories and only recruit from degree-granting programs; internship pipelines are almost entirely degree-dependent
• Several Make School graduates without degrees have landed at Google and Facebook — but these are individual exceptions, not structural pathways
• The structural problem: all 100 IPOs in a Silicon Valley boom cycle transfer wealth primarily to people from the top 20% of families
• Without structural access routes, anecdotes of low-income success in tech do not move the equity needle

The future of higher education

• Make School is among the first colleges created under WASC's 2014 accreditation rules — potentially the start of a new wave of institutional entrants
• Traditional colleges will face increasing pressure to align incentives with students as the student debt crisis becomes more visible to parents
• Purdue and other established schools are already adopting ISA frameworks using documents co-authored by Make School in 2014
• Harvard and similar schools will become more openly understood as luxury goods; outcome-focused alternatives will become more legible
• Expect students to enter the workforce earlier and return for targeted education later — lifelong learning in cycles rather than degree-stacking
• Many college programs will close as their ROI becomes undeniable; a higher-ed bubble correction is likely

Advice for YC founders

• The anti-pattern is assuming you are above foundational YC advice: make something people want, focus, do not do fake work
• Skip conferences and coffee meetings unless they directly affect revenue
• Track default alive vs. default dead weekly — strategic awareness of your runway is not optional
• More startups than expected come within weeks of running out of money through avoidable sequencing mistakes
• Optimising fundraise terms matters far less than not dying; shift from default dead to default alive first