One moment
Retrieving
Three kinds of mission, and everything the Bureau has published so far. Every one is worked out on paper and checked on screen, and every one is marked as it goes.
An Operation is a complete escape room sized for a single lesson. Locks open in order, each one gating the next, and the last one wants the code the others have been handing out a digit at a time.
Five locks stand between a class and a jammed Bureau archive: place value in the millions, rounding to find a signal among decoys, negative numbers across zero, and Roman numerals. Each lock yields one digit of the restore code, and the code is the way out.
A sealed jar reaches the Bureau’s lab overnight with no paperwork: pond water, sediment, and a photograph cut into six pieces. Five tests on the bench, in any order: read a heating chart, rank five tapped tubes by their note, weigh the dried sediment, test what conducts, and key out what was living in the water. Every result goes on the lightbox record. When all five are logged the analyser opens, and the photograph, put back together, says where the jar was filled and who is clearing the site.
A sold-out comeback is hours from going live and somebody inside the record label is quietly ruining it. The board is the building’s camera wall: six rooms, ten pieces of tampered material to work through in any order (lyric sheets, a press release, the fan wall, interview captions), each one turning on word classes, figurative language or the exact shade of a synonym. Ten findings patch the wall; the last feed asks who did it.
The Bureau keeps a tin of small change for the Friday market, and this Friday it does not balance. Nobody has taken anything. Work five stalls on paper (loaves at a price each, trays of eggs, two cheeses added in columns, the change from a bag of apples, roses and posies), and type what each one should really have cost. Every total finds its row on the Stall Code and gives a letter; the tin gives the sixth; and six letters, typed at the brass scales, name the thing that was wrong. Year 3 arithmetic, one instruction a sentence, and nobody is blamed.
Four of the 23 Operations in the field. Every one of them is in the catalogue once you sign in.
A Case runs over several lessons, and each phase targets a different part of the curriculum. The work accumulates: what a class finds in phase two is what they are ranking in phase five.
Students learn how the world's economies actually compare while they solve the puzzles: a six-phase investigation that ends with a card deck their class can play with.
The Global Economics Bureau needs field intelligence. Each agent picks four to six countries out of a real 61-nation dataset, under constraints that force a varied list: one nation over 100 million people, one under 10 million, one with GDP per capita over $50,000, one under $5,000. Everything they calculate from then on is about the countries they chose, so no two decks in the room are the same.
Agent Training is not authored the way an Operation is. It runs off a question bank (one you write, or one from the Bureau Library), and a bank built for one mode runs in every other one without being rebuilt.
One question to the whole class at once, on a shared screen, with everyone answering on their own device. Typed answers, not just multiple choice, so a child writing 282 has done the multiplication rather than picked it out of four. No setup and no accounts: you read the PIN off the board and the class is in.
The same bank, played alone, at whatever speed a child works at. Useful for the ones who want another go at a bank the class has already played.
The Bureau is early, and every mission here was built for a real classroom before it was published. If there is a unit you would want one of these built around, that is a useful thing for us to know.
A Top-Trumps-style intelligence deck, built from figures they worked out themselves rather than copied. The maths is the mission, and the deck is the reason it mattered.
The board turns the question over: the right answer, the explanation the bank carries, and how the class split across the options. A teacher can see which wrong answer was chosen rather than only how many got it, and that is the part worth talking about before the next question.
No child is named on it. The counts are aggregates, and a typed answer never appears on the shared screen at all.