Two delivery partners on what an eleven-million-dollar education bond actually asked of the classroom.

Pankaj's argument is sharper than the standard NGO line about "the teacher problem". He says explicitly that mainstream education thinkers have spent twenty years opposing structured pedagogy on the grounds that it de-skills the teacher. His response: the average primary-school teacher in rural Gujarat is not being de-skilled; she is being respected. The system being asked of her — to behave like a four-year-trained college professor managing a forty-child mixed-ability classroom — is the de-skilling, because it is impossible.

The Gyan Shala design starts from the actual skill level and works backwards. Lesson scripts are pre-structured. Materials are pre-printed. The three interactions are pre-engineered. The teacher does not need to be the curriculum designer; she needs to be a competent executor of a curriculum designed by someone who understands how a child's mind acquires language, deductive thought, and inductive thought.

The unusual claim. Pankaj says a child can achieve mastery of grade-level learning in three to four hours a day, in a classroom that is non-threatening and non-paralysing, with the three interactions running properly. Gyan Shala runs in that frame: a focused supplementary or parallel session that the slum child can fit alongside whatever else her day demands.

The implication for policy is uncomfortable. Indian government schools run six to seven hours; private schools run longer. If three to four well-designed hours match or exceed the outcomes of six to seven undesigned hours, the question is not how much instruction time exists but how it is being used. The conversation does not labour the point. It does not have to.

Vishal's framing question was the standard one: poor families pull children out of school because they need them to earn. Pankaj reframes it. In his read of Gyan Shala's experience, ninety percent of dropouts are not earning-pressure — they are children who concluded the school is not giving them enough to stay. The five-to-ten percent who drop out for economic or social reasons are real, but they are the minority case.

The implication is that drop-out is downstream of learning outcomes, not upstream of them. Pay families to keep children in unhelpful schools and the children still leave; build a school in which children visibly learn and the families do not need paying. The lever is the classroom, not the cash transfer.

The most precise small moment in the conversation. Vishal asks how to make school fun; Pankaj declines the word. Fun risks becoming the objective, and an objective of fun is satisfied by sweets and games, not by learning. Pleasant is the right frame, and the pleasure that retains children is specifically the pleasure of mastery — the unmistakable internal signal that you have learnt something.

The distinction is not pedantic. Twenty years of edtech in India has been built on the fun frame and largely failed. Prem says explicitly later in the conversation that the assessment-and-grading edtech company he saw fail "very weird man" was operating in that frame. Pleasure-of-mastery is what carries a child back into the room the next day.

This is the most operational moment in Pankaj's section. Inside a class period, exactly three interactions happen: child with teacher, child with other children, child with the learning material. He calls them out as a designable triple and notes that all three have to be engineered in coordination, not in isolation.

Most reform programmes obsess about the first edge (teacher-training initiatives), occasionally touch the third (better textbooks), and almost never touch the second (peer interaction). Gyan Shala's classroom design is unusual in that it loads all three on purpose, with the second edge contributing a non-trivial share of the outcome. The Class 5 child does not learn only from the teacher; she learns from the Class 5 child next to her, in a structured peer arrangement that the curriculum designer has set up.

Pankaj pushes back when Vishal compresses his model into "deductive learning". His correction is patient. Language is its own thing — children acquire language; they do not learn it the way they learn mathematics. Mathematics is most cleanly the home of deductive thinking. Science is the home of inductive thinking. The classroom designer has to understand these three different ways the human mind engages with knowledge, and design class processes accordingly.

The seemingly academic distinction is structural. A class that teaches mathematics with the cadence of a language class will fail; a class that teaches a language with the cadence of a science class will fail differently. Gyan Shala's curriculum is built on the recognition that these are three different operations on the mind, and that an effective classroom switches between them deliberately.

Prem's number is exact. The Annual Status of Education Report — Pratham Education Foundation's flagship survey, running for fifteen years at the time of recording — puts primary-school enrolment at 98.5%. India has solved the access problem that consumed two decades of policy attention. The companion number, also from ASER, is that fewer than half of Class 5 children in a government school can read a Class 2 text. Five years of schooling produces a child who is, in literacy terms, three years behind.

The diagnostic implication is that primary education in India is no longer an enrolment problem and is now a learning-outcomes problem. The two require different instruments. Enrolment was moved by mid-day meals, school construction, RTE-style entitlements. Learning will be moved, if at all, by classroom design, adaptive placement, and outcomes-based monitoring. The DIB is a first instrument calibrated for the second war.

Both Pankaj and Prem refuse the leading question. Vishal opens by asking how the Quality Education India DIB changed the way they teach. Pankaj corrects: a bond is a financing mechanism. It has nothing to do with the education per se. What worked in a classroom worked whether it was funded by a DIB, a grant, or a government cheque; the design of the learning process did not change because the funding structure did.

What did change was governance. The DIB locked the contractual learning outcomes to a negotiated subset of the state curriculum, formalised process monitoring that had previously been informal, and required testing to be done by a third party with its own framework. None of this is pedagogy. All of it is overhead, sometimes useful, sometimes costly. The intellectual honesty here is rare — outcomes-based finance is often sold as a pedagogy transformer, and it is not.

An overlooked detail. The DIB structure required Gyan Shala's progress to be benchmarked against a comparison group of similar socioeconomic profile. The ideal comparator would have been a government school. The government — Pankaj says with diplomatic understatement — would not agree to be the benchmark. The comparator that was negotiated was a set of low-cost private schools serving children of broadly similar background.

This is two findings, not one. First, the government's reluctance is itself evidence of where accountability for outcomes sits in the Indian primary-education system: nowhere stable. Second, the comparator that was chosen is the harder one — low-cost private schools are paid for by aspirational parents and have stronger incentives than government schools to produce visible literacy. Gyan Shala outperforming the comparator by two-to-three-times the contracted target is therefore the stronger version of the result, not the weaker one.

A second overlooked detail of the bond's design. The independent tester chosen by the British Asian Trust did not run a literacy test (read these letters, decode these sounds). It ran a comprehension test (read this passage, understand it, answer questions on it). Pankaj is clear that this is a stronger test, and that the Class 5 children failing the ASER literacy test would have failed this one more severely.

The choice mattered. A test of comprehension forces the curriculum to teach for comprehension; a test of literacy permits the curriculum to teach for decoding. Gyan Shala's outperformance on the comprehension test is a stronger claim than the same outperformance on a literacy test would have been, because comprehension is harder to game with surface drilling. The DIB structure, by handing the testing to a third party with sophisticated instruments, raised the bar for the delivery partner and made the result more credible.

Pratham InfoTech's Lucknow build is the most operationally specific stretch of the conversation. Fifty-five Uttar Pradesh government schools, twelve hundred Chromebooks, a learning lab per school with twenty-to-twenty-five computers in one classroom, MindSpark adaptive software underneath. Fifty-five further schools chosen as a matched control. The treatment was a daily target — thirty-five minutes of MindSpark per child per day. They called it "Mission 35." A schedule that took weeks of negotiation with principals and management to fit into the school day.

The daily-data discipline was the hidden lever. Because MindSpark logged every interaction, Pratham could rank schools by daily performance, identify a top five and a bottom five, and propagate the practices of the top five to the rest on the same day's cycle. That feedback loop was tighter than anything paper-based monitoring could have produced. The treatment group ended up at a forty-three-to-forty-five-scale-point baseline versus the control at nine. The fivefold lift is the headline; the daily ranking-and-propagation is the mechanism.

Prem's clearest framing of what the Pratham-family methodology has always called Teaching at the Right Level. Imagine a Class 5 student sitting in a Class 5 lesson while reading at Class 2 — she is lost, and the teacher's instruction is irrelevant to her actual cognitive state. The model has to start her at Class 2, and route her up one rung at a time, regardless of which standard she is enrolled in.

On scale, the math is grim. India has roughly three hundred million children in school. If half are lagging — which is what the ASER reading numbers say — that is one hundred and fifty million children sitting in lessons they cannot follow. Adaptive placement, whether done by a teacher in a small-group setting (Pratham Education's original Read India methodology) or by software at thirty-five minutes a day (MindSpark in the Lucknow build), is the only mechanism the conversation describes that scales without adding teachers India does not have.

The most underappreciated data point in the conversation. When the lockdown hit in March 2020 and Pratham InfoTech wanted to move its Lucknow programme to remote delivery, the first thing they did was count. Of twelve thousand enrolled students, twenty-two percent of families had a mobile phone, and only twelve to thirteen percent had a smartphone. Eight months later, the smartphone number was at seventy percent. Eight months. Among government-school families in Uttar Pradesh.

The two readings of this are different. The optimistic reading is that the parents — government-school parents, not affluent ones — invested aggressively in the child's education because they understood it mattered, and reached for the only available delivery rail. The pessimistic reading is that the delivery rail collapsed pedagogically anyway because the children dissociated from a year of fragmented online instruction. Prem holds both readings simultaneously. The smartphone exists; the school still has not figured out how to use it.

Pankaj's flattest line in the conversation. Asked whether India has caught up on girl-child education, he says no — and explains. Bangladesh is now at per-capita-income parity with India, largely because it solved girl-child education twenty-five years ago. India has spent more per capita on education than Bangladesh for the same period. The result is that India still does not have parity, and Bangladesh does.

The implication for anyone who treats education spending as the policy lever is unwelcome. Allocation is not the binding constraint. Conversion is. Money spent on a model that does not produce results produces no results, regardless of the amount. The Bangladesh story should appear in every Indian primary-education debate as a permanent footnote: spending more was not the missing thing.

The closing exchange is the most honest stretch of the conversation. Vishal asks whether India needs more DIBs. Pankaj says yes in principle — paying on outcome is fundamentally desirable. Then he says no, in practice, at scale. Testing on a hundred-times-larger sample is itself a complex and expensive operation; the negotiation overhead of a structure with this many stakeholders is high; and the cost-per-outcome of running another DIB at one hundred times the size would be a different ball game.

Prem agrees. The instrument worked. The design needs to be simpler and the testing needs to be cheaper before another one can run at the scale India actually needs. Both are careful to thank the British Asian Trust for starting target-based financing in Indian education. Both make it clear that the next instrument must inherit the principle, not the structure.