The AmplifierReal cases · Education · Teacher
Farah, 31: forty-five students, five levels, one lesson plan
A class where a third are lost, a third are bored, and the middle is the only group the textbook was written for.
Planned for release phase 2. The page is complete as a draft; interactive parts run in the browser and do not yet save to an account.
In brief
The situation. Farah teaches Grade 8 science at a school in Lucknow. Forty-five students, a syllabus that assumes one speed, and a term that ends in a board-pattern exam. She knows her class has roughly five levels in it and has been teaching to the middle for six years because there was no other way. The school has sent a circular about "integrating AI in the classroom" that describes nothing she can use on Monday.
Step 1 — Define the End State
"In the photosynthesis unit, the eight students who are lost can explain it to me by week three; the six who are bored have done something harder that they chose; the middle keeps its pace. I spend my evenings on the three students who need me, not on making five worksheets. And the exam results do not drop." Audiences: three named groups of students, the principal, and herself at 9 p.m.
Step 2 — Map the Flow
Her list: a quick diagnostic per unit, three versions of each worksheet, an extension task for the top group, a way to check who moved. AI review adds: a criterion for "lost" that is not a mark — a misconception list per topic, because the same wrong answer has different causes; and a rotation so the top group also does some teaching. It suggests an adaptive-learning platform; the school cannot afford it and she says so.
Step 3 — Assign Tasks
The misconception list: hers first — six years of the same wrong answers — then AI adds what she has not seen. The diagnostic: AI drafts, she edits. Three versions of each worksheet: AI generates from her single master; she checks every question. The extension task: AI generates options; the top students choose. Explaining photosynthesis to the eight: her, in person, no tool.
Step 4 — Select the Tool
A conversational tool with a project holding her syllabus and her misconception list, used weekly. She never puts a student's name into it.
Step 5 — Prepare and Direct
Criteria: every worksheet question maps to one misconception on my list; the three versions differ in scaffolding, not in content; the extension task cannot be finished by searching; anything I hand a student, I have solved myself first.
Steps 6 and 7 — Execute and Evaluate
The first set of three worksheets fails: the "easy" version is shorter, not more scaffolded — the tool had confused difficulty with length. She rewrites the brief with an example of scaffolding from her own hand. The extension tasks are good; two are searchable and fail. The diagnostic works and tells her something new: three of the "lost" eight share one misconception she had never named.
Step 8 — Integrate and Stress-Test
Week three. Small fix: the top group's rotation clashes with lab days. Structural: the middle group's pace has slowed — because she has been spending class time with the eight, and the worksheets were never meant to replace her for the middle. A Step 2 flow gap: her plan had no task for what the middle does while she is with the eight. Added.
What changed
Seven of the eight can explain photosynthesis by week three. The results hold. Her evenings are shorter. The circular about AI is still on the noticeboard; her version of it is a misconception list and three worksheets.
A misconception list is a teacher's end state in disguise. "I have solved it myself first" is the criterion that keeps the tool honest. Scaffolding is not shortness — give the tool an example in your own hand.
A composite case, not a client engagement; the person is invented, the situation is not.