The Cabinet Secretary comes in last, carrying nothing but a pen. The sheet that has lain on the desk since 21:07 is where the night left it: forty lines under one word, TRIAGE, sixteen of them stamped. The Cabinet Secretary writes the folders’ remaining orders into the other twenty-four, each beside the scene that earned it, until every line carries something.
Triage sheet: 40 of 40 lines filled“Forty,” the Cabinet Secretary says. “A morning can carry three. So we choose, and we choose by three questions. What does it cost? Are the people it hurts many and quiet, or few and organised? Does it need an institution that does not exist yet?”
Line by line the pen goes down the sheet. Thirty-seven lines are struck. None is abandoned: they are struck from tonight’s speech, not from the plan, and each is moved to a column of its own.
Triage: this book's ranking, one defensible ranking among several
- The first three sentences. High legitimacy, low cost, long horizon: a scoreboard with independent statistics; a compensation-before-pain compact with the states; every child reading by Standard III.
- The first budget. An evaluation unit; tribunal capacity; a fiscal anchor that includes the states; power subsidies paid in advance, by transfer.
- Build quietly. A published tariff glide path; sunsets on quality orders; land-record audits; fallbacks for digital systems.
- Postpone until evaluated. Changes to labour thresholds beyond the codes; decoupled farm payments beyond pilots; any new large incentive scheme.
The three that survive are not the ones you would have picked at 21:00. They are the ones that cost least to say and cannot be done later. A scoreboard defined after the scores are in is not a scoreboard. A compensation promise made after the pain has landed is read as a bribe. And a child who learns to read pays back slowest of all, so that one has to start first.
“Here is the ruler,” the Cabinet Secretary says, and puts one last page on top. It is the scoreboard, with nothing on it that is a target.
The scoreboard: baselines, not targets
- Income per head, constant 2015 dollars. $2,523 in 2025.
- Workers in agriculture. 41.6 percent in 2025.
- Manufacturing, share of GDP. 13.47 percent in 2025, down from 17.03 in 2010.
- Standard III government-school children who can read a Standard II text. 23.4 percent in 2024, rural.
- Insolvency resolution. 713 days on average, against a 330-day limit.
- District courts. 5,16,39,674 cases pending, of which 47,02,876 are over ten years old; the snapshot is dated 21 October 2026.
- Power lost between plant and meter. 14.5 percent of output in 2024.
- Cities’ own revenue. Under 0.6 percent of GDP.
- Public finance. The Centre’s debt is 56.1 percent of GDP, with a target of 50 percent, give or take one point, by 2030-31. Interest takes about 36.8 percent of the Centre’s net revenue; the states’ liabilities are about 28 percent of GDP.
- Statistical credibility. The IMF grades India’s national accounts C, the second-lowest of four.
Beneath the table is the faint pencil line the Statistician’s sheet ruled forward from 2025 at the start of the night, carried over unchanged, and under it two words: ARITHMETIC, NOT FORECAST.
The state finance minister is not on any schedule. The minister’s party sits across the aisle from yours, whichever yours is, and the minister has come in person because the question will not go by messenger. The Room has no names, so this one is an office too.
“You have three sentences,” the minister says. “One of them had better be about us. What do the states get?”
You answer with what the night has put on the desk. Chapter 23 told how the Goods and Services Tax passed: the states were given two-thirds of the weighted votes in the Council that sets its rates and a written promise of five years’ compensation. The promise ran out in June 2022 and nothing took its place.
“I know how it ended,” the minister says.
Nobody in this book needs to be called the villain of the story, and you do not make the minister one. The party that led resistance to the bill in 2011 carried the tax through in 2017; the side that blocks changes with who holds the Centre. That is the reason to write the terms down.
“A written promise to whom, for how long, and paid from what?” the minister asks. You do not name a sum. The sum is for the Finance Commission and the Council to settle with the states in the room, and a number you invent at 05:40 would be read as a bargaining position. You commit to three things.
First, a compensation fund, set up and funded before any change to a rate or a tax base, so that no state learns the price after the bill passes. Second, conditions through the Finance Commission, which already works by formula and whose award runs from 2026-27 to 2030-31. Third, the fairness test, said aloud: would you accept this power in an opponent’s hands? In April, seats and the women’s quota travelled in one bill, and the bill failed.
“Then say the test out loud,” the minister says, “with the light on.”
“I will.”
The choice that hurts is that you have just promised money you cannot yet price to the very people who will cost you the most votes. You make it anyway, because the other course is the one the record already shows.
The Cabinet Secretary spreads the last sheet along the desk’s length. It is a strip, ruled in bands, with four bands across the top and the dates already set by notification, award or announcement pinned along the bottom.
The calendar strip
- Years 0 to 1, build the machine. Most of it costs little: statistics and the scoreboard, the evaluation unit, tariff neutrality and sunsets, state labour rules, tribunal capacity, digital fallbacks, a fiscal frame that includes the states.
- Years 1 to 3, prove it in a few places. Six to eight clusters and five to ten city-regions where land, power, labour rules, credit, transport, hostels and creches are fixed together; discom dashboards; a daily learning hour; primary-care audits.
- Years 3 to 7, scale what the evaluations support. Nothing above a size threshold scales before an independent evaluation reports, and nothing is cancelled before one does.
- Years 7 to 10, deepen. Components and capital goods; storage-backed decarbonisation; renegotiated trade deals.
- Set, 31 October 2026. States are expected to have notified their rules under the four labour codes.
- Set, 1 March 2027. The census reference date.
- Set, 31 March 2031. The sixteenth Finance Commission’s award ends. The Centre’s debt target is due by 2030-31, and the Reserve Bank’s 4 percent inflation target is set to run until 2031.
The book leaves the year you took the oath blank, so the four bands slide with it. The dates do not: whatever year you begin, they are what the calendar already holds, and where one has passed, it is behind you. The plan assumes no one in particular and no election.
You ask what the calendar is for. “So that your opponents can read it,” the Cabinet Secretary says. “If they can read it, they can tell the country when you are late.”
There is one paper on the desk that is not stamped 21 OCT 2026. It is in an unmarked folder, the only one tonight without an office on the tab. Nobody in the room claims it, and you do not ask.
Inside is a photocopy of the Xiaogang contract of 1978, red prints, seals and the pledge to raise the cadres’ children: the paper the prologue watched being made.
Behind it is a second sheet, a letter from Tembhli, in Maharashtra, where on 29 September 2010 Manmohan Singh launched the Aadhaar number. Chapter 22 tells what came of it.
You hold them side by side. One is a defiance; the other is a registration. They are not the same act and the book has never said they were. Both are a poor person’s mark.
The Cabinet Secretary has one more page, and it is a shape, not a speech. It is the shape of the last time a Prime Minister of India spoke to the whole country in a crisis and asked it to carry a cost.
Chapter 10 heard the broadcast; what the desk needs from it is the order. On 9 July 1991 the Prime Minister told the country that the repair would not be easy, quick, or smooth
, and in the same speech named the crisis, promised the poor protection and credited a predecessor. Naming the crisis, tying the pain to a gain and promising the poor you can do tonight. The credit is a choice. Rao credited a predecessor of his own party; the version a relay needs credits someone from the other side, so that the other side can finish it.
“Credit goes to whoever finished it,” the Cabinet Secretary says. “That is how the other side comes to finish it.”
You think of a schoolteacher in Kalpa. Shyam Saran Negi’s ticks run from the first vote of independent India, on 25 October 1951, which A Basement watched, to his thirty-fourth, on 2 November 2022, which A Blank Sheet of Paper recorded. The next ones are not his.
The ledger is on the desk. It is a plain book, ruled, with a first line that has not been filled. Every order from tonight will be written in it with its test, its date and its cost, and anyone will be able to open it. You are handed a pen.
There is no ink pad. There is no thumbprint. You sign the first line with the pen, and the reader is left to make the rhyme.
The red light comes on.
The address, as you give it:
I will say three things this morning, and I ask you to hold me to them.
The first is the ruler. From today, the numbers by which this government may be judged are published, along with how each is measured, and they do not change while I hold this office. Income per head, in constant dollars, was $2,523 in 2025. That is where we start. The office that counts must answer to a law and not to me; it will release its raw data, and it will never suspend a series because the series is inconvenient. India’s national accounts are graded C by the International Monetary Fund. I would like a better grade, and I would like you to be able to check.
Replay of an earlier scene: The scoreboard, ch 1The second is a promise to the states. Before any change that costs a state revenue, or a worker a job, the compensation will be written down, funded, and known: by place as well as by sector. Research on India found that after trade was opened in 1991 the rural districts most exposed to it saw slower falls in poverty. The fund set up then was aimed at industrial workers: voluntary retirement in public enterprises, training and counselling. This time the promise comes first. The Council that settled the tax shows how: the states that stood to lose were given most of the votes and a promise in writing. The promise ended in 2022. I will not let the next one lapse without a successor.
Replay of an earlier scene: The Council, ch 23The third is a child. In 2024, 23.4 percent of Standard III children in rural government schools could read a Standard II text. It was 16.3 percent in 2022. It is better, and it is far from enough. Every child reading by Standard III is the one promise I make that you will not be able to check before the next election. It takes years to show. I make it anyway.
Replay of an earlier scene: The report card, ch 27I am not telling you the other thirty-seven. They are on a sheet, in the open, with the first budget, the things we build quietly, and the things we will not do until they have been tested. I do not say them tonight because a speech can carry three things, and I would sooner you remembered three.
Here is how I intend to work. Nothing large will scale before an independent evaluation reports, and nothing will be cancelled before one does. In November 1978, in a village in Anhui, eighteen households pressed their marks to a contract that, if it failed, could have cost their cadres their freedom or their lives. By 1980 Deng Xiaoping was pointing to counties where household contracting had been allowed and output had risen. A trial that is protected, and measured, can be scaled. I ask my own ministers to give states the same room.
Replay of an earlier scene: Xiaogang, 24 November 1978You will hear that I am wrong, and some of the people who say so deserve a hearing. Raghuram Rajan and Rohit Lamba argue that the world lacks the consumers for another China-sized manufacturing economy, and that subsidising factories carries a cost. Jean Drèze and Amartya Sen counted the Chinese famine’s deaths against India’s absence of large famines, and asked what a growth rate leaves out. Jagdish Bhagwati and Arvind Panagariya argue that only market-led growth helps the poor to any significant degree. I do not think those arguments are finished, and I will not pretend they are. The scoreboard I have published includes the things on which each side would judge me.
There is also a view of land that is not about price. Compensation near market value was not enough in Singur: land was misclassified, the incomes of tenants and owner-cultivators fell, and most households preferred something other than cash. The law of 2013 asks for the consent of eighty percent of those affected by a private project. The ordinance that followed it in 2014 was meant to make acquisition less onerous. I will audit the records before I ask anyone to give up land, and I will ask by consent.
Over these years, measures have been turned back under different governments. In 2008 spectrum was handed out first come, first served, and in 2012 the Supreme Court cancelled 122 of those licences. A tax was made retrospective in 2012 and nullified in 2021. A land Bill that exempted five categories from consent passed the Lok Sabha in March 2015 and did not become law. Three farm laws were repealed by a bill introduced on 29 November 2021. A constitutional amendment on seats and the women’s quota fell in April of this year. These are not one side’s failures. Each is what happens when a change is made faster than the people it touches can be consulted.
So I will ask every power I seek the question I asked the minister in my office this morning: would I accept it in my opponent’s hands? If I would not, I will not ask for it. This matters for a plain reason. Funke, Schularick and Trebesch studied fifty-one populist leaders between 1900 and 2020, on the left and on the right, and found income per head about ten percent lower fifteen years later than it would otherwise have been. It does not matter which flag.
And I will not promise you a miracle. Since the 1990s only 34 middle-income economies have become high-income, and 108 were still in the middle at the end of 2023. If India grows at 5.7 percent a year per head, as it did in its best decade, income reaches about $4,390 in ten years; at the pace Korea held from 1987 to 1997, about $5,250. That is arithmetic, not a forecast. I do not know which of those we will get. I know the dates on which you can ask: 31 October for the states’ labour rules, 1 March 2027 for the census, and the end of March 2031 for the award that settles what the states receive.
There is a ledger on my desk. I have signed its first line, and anyone may open it. I am asking you for ten years, and I am asking you to read it.
The light goes off. On the desk, beside the signed ledger, the bottle that has stood there since 21:00 is still unopened.