Public education systems are very good at reaching scale and much less good at holding quality while doing it. Almost every large technology education initiative meets the same five constraints, in roughly the same order, and the ones that succeed are the ones that designed for those constraints rather than discovering them in year two.
Constraint 1: The classroom does not run in the curriculum’s language
This is first because it is the one most often treated last. A programme written in English, delivered to a classroom operating in Marathi or Bahasa Indonesia, imposes a translation tax on every session — paid by the teacher, live, while also teaching.
Designing for this means the language layer reaches teacher material, not only student material. A translated student workbook with English lesson notes is not a multilingual programme; it is a multilingual veneer over a monolingual one.
Constraint 2: There are not enough specialists, and there will not be
Any plan that depends on recruiting technology specialists at national scale is a plan to fail slowly. The arithmetic does not work in any large system.
The alternative is to design the programme so a capable non-specialist can deliver it well: worked reference solutions, anticipated misconceptions, explicit session structure, and a defined fallback when the equipment does not cooperate. This is a design constraint that improves the programme for everyone, including the specialists.
Constraint 3: Infrastructure varies enormously within a single district
Two schools ten kilometres apart can differ by an order of magnitude in devices, connectivity and usable lab time. A programme with one delivery assumption will work in a subset and be quietly abandoned in the rest.
The useful distinction is between the standard and the delivery. The standard — what a student should be able to do at the end of the year — should not move. The delivery — shared devices, rotation, offline-capable material, unplugged sequences for concepts that do not strictly need a machine — should be configurable per school.
Constraint 4: Programmes disappear from view between launch and audit
A large initiative is announced, funded and launched. Then, for eighteen months, the only available information is that it exists. The next data point is an evaluation, at which point the correctable problems have compounded.
The fix is unglamorous: visibility from classroom level upward, on things that are actually informative. Not attendance alone — participation, project completion, teacher activity, and where cohorts are stalling. The purpose is not surveillance of teachers. It is knowing which districts need support before the year ends rather than after.
Constraint 5: Nothing survives the initiative
The most demoralising pattern in public education technology is the three-year cycle: an initiative arrives with hardware and training, runs, ends, and leaves a store cupboard.
Durability comes from what stays: curriculum owned by the system, teachers who are permanently more capable, material that does not expire with a licence, and a progression that continues into the next grade whether or not the programme is re-announced.
What a defensible rollout looks like
Enable before you deploy. Teacher training precedes student sessions, without exception. A deployment that reaches classrooms before its teachers are ready spends its first year manufacturing scepticism among the people it most needs.
Pilot with a real definition of success. A defined set of schools, a defined grade band, a defined term, and — critically — agreed criteria written down before it starts. Pilots without prior success criteria always succeed, which is why they are worthless.
Instrument from day one. Retrofitting monitoring onto a running programme rarely happens.
Expand on evidence, not on schedule. The most valuable output of a pilot is the list of things that did not work. A rollout plan that cannot absorb that list is a schedule, not a plan.
Plan the second cohort before the first finishes. Grade 6 students become Grade 7 students. Programmes that only fund an entry year create a cliff.
What scale actually buys
The case for public deployment is not only equity, although that is sufficient on its own. It is capability distribution.
The problems these students will inherit — water, agriculture, climate, health systems, public infrastructure — are distributed problems, and disproportionately concentrated in exactly the districts that public education serves. The talent to work on them is distributed too. A system that develops technical capability only in its best-resourced schools is choosing to leave most of its own potential undeveloped.
That is not primarily a fairness argument. It is a capacity argument, and it is the strongest one available.