The coursework is already free.
Almost nobody finishes it alone.
Harvard and MIT gave their courses away. Only 3.13% completed the courses.
Researchers at MIT then analyzed every course the two universities delivered on edX between October 2012 and May 2018 — the most complete dataset that exists on self-directed online learning. Three numbers carry the argument, and a fourth explains why the first three did not fix themselves.
The number people already half-know. It is usually explained away — the material is too hard, or the people who signed up were never serious. Neither explanation survives the next number.
Learners who paid to enter the verified certificate track took the same courses. They were not given easier material, more time, or better instruction. What separated them is that they had committed, and something was at stake.
The failure is not learners struggling partway through hard material and giving up. Whatever gets someone from intending to learn to actually starting is missing — and no amount of course quality supplies it, because the learner never reaches the course.
Completion is a function of accountability, not of content quality or content access.
This is the part most often missed. Across those years the platforms improved on every dimension a product team would measure, and completion went down. The problem is not an interface problem and it will not be solved by an interface. Anything built on the assumption that better delivery fixes completion is building against six years of contrary evidence.
Harvard and MIT gave all three away. It didn’t work.
Self-directed completion depends on conditions that are unevenly distributed: uninterrupted time, a quiet place to work, a peer who has done it before, someone to ask when the material stops making sense, and the confidence that finishing will lead somewhere. A learner working thirty hours a week has fewer of these than a learner with a graduate degree taking a course for professional interest. The same free course, offered to both, produces very different outcomes — so free access on its own widens that gap rather than closing it. These are the populations ALEX serves.
The program is named for the problem. ALEX Institute supplies the layer the evidence says is missing, and only that layer.
Assigned by name. Makes contact before the coursework starts, checks in on a schedule, and follows up when a learner goes quiet. This is the accountability that paid enrollment created for verified learners — supplied at no cost to the learner.
For each class or section. A practitioner who can answer the question that stops a learner cold. In self-directed study, one unanswered question is frequently where completion ends.
Neither of these is course content. ALEX writes no curriculum, delivers no lectures, and issues no learner credentials. The providers supply the coursework and their own credentials; ALEX supplies the human structure around them. Free to students. Every student. The course flow overleaf shows exactly what a learner takes, and from whom.
The gap is someone expecting you to show up.
Cohorts begin November 2026. Volunteers are needed per student and per section, not per program — one mentor is one completed course that would not otherwise have happened.
Take one of these
Or refer someone — as many people as you like, in any of the four roles. Name, email and phone for each; we contact them.
| Claimed | Structured human support materially raises completion relative to unsupported self-directed study. This is well established across the literature. |
|---|---|
| Not claimed | That ALEX will reach 46 percent, or any particular figure. The beta target is 40 to 60 percent cohort completion — stated as a range because it has not yet been tested against real cohorts. |
| Not used | The 85 percent completion figure reported by South Texas College. That result was driven by a course-grade requirement rather than coaching, which makes it structurally different from this program and inappropriate as a benchmark for it. |
| Two measures | Cohort completion and credential attainment are different numbers. Several provider credentials carry assessment thresholds that sit outside ALEX’s control — a learner who attends every session, does every assignment and misses a quiz threshold has not failed the ALEX program. Both figures should be reported; the 40–60 percent target applies to cohort completion. |
The 3.13 and 46 percent figures come from a single well-constructed study of one platform, which is a strength — the comparison is internally consistent and no cross-study methodology problems apply. It also bounds the claim: edX learners are not identical to the populations ALEX serves, and MIT and Harvard courses are not identical to Google, IBM and NVIDIA certificate content. The defensible claim is the pattern — structure and accountability drive completion far more than content or access — not the transfer of either number to an ALEX cohort.
Source. Reich, Justin, and José A. Ruipérez-Valiente. “The MOOC Pivot.” Science 363, no. 6423 (2019): 130–131. doi.org/10.1126/science.aav7958 — all MIT and Harvard courses delivered on edX, October 2012 through May 2018. “Verified” learners are those who paid to enter the certificate track; “completers” are those who achieved a passing grade. Learner counts: Class Central annual MOOC census, 2015–2018.