Taleemabad Field Intelligence · Issue 06 · May 2026

We stopped asking teachers "how did it go?"

Good coaching is the highest-leverage thing that can happen in a classroom — and the rarest. Most teachers on Earth are observed once every few years, if ever. So we gave every teacher a coach in their pocket. The hard part wasn't the listening. It was the asking.

LIVE IN PRODUCTION · PK · TZ

A great coach never asks a question you can answer on autopilot. For months, ours did. It listened to a recording of a teacher's lesson and then asked the safe, generic things — How did the lesson go? What would you change next time? — questions a teacher learns to pattern-match by her third session.

The answers showed it. They were getting shorter. The share of teachers who walked away with a concrete "next time I will…" plan was sliding. A coach that asks forgettable questions gets forgettable answers — at any scale.

So in May 2026 we rebuilt the question engine from the ground up, around a single rule: never ask a question the teacher could have answered without having taught that exact lesson.

A generic question gets a generic answer. We wanted the AI to put the teacher back inside the one minute of her lesson that actually mattered. — Design principle · Coaching v12
One moment · three ways · zero templates

The same 30 seconds, asked three different ways

Ehtisham's maths lesson, 20 May. The AI noticed a contradiction in his own teaching: at 05:00 the children were counting on their fingers — discovering it themselves. By 11:20 he'd handed them a rule: "always start from the unit." Asked to write a reflective question about that turn, the engine wrote three genuinely different ones. Each arrives as a WhatsApp voice note, in Urdu.

Variation A · the counterfactual
0:18
05:00 پر بچے لائنز کھینچ کے گن رہے تھے — اپنے ہاتھوں سے سمجھ رہے تھے۔ پھر 11:20 پر اچانک ایک رول آیا کہ ہمیشہ یونٹ سے سٹارٹ کرو۔ ان دو لمحوں کے بیچ میں کیا ہوا جو آپ نے رخ بدلا؟
"At 05:00 the children were counting with their hands — understanding it themselves. Then at 11:20 a rule appeared: always start from the unit. What happened between those two moments that made you change direction?"
Variation B · the learner's mind
0:15
ایک ہی لیسن میں دو الگ طریقے — 05:00 پر "خود گنو" اور 11:20 پر "رول یاد رکھو"۔ آپ کے خیال میں بچے کے ذہن میں یہ شفٹ کیسی محسوس ہوئی ہو گی؟
"Two methods in one lesson — 'count it yourself' at 05:00 and 'remember the rule' at 11:20. How do you think that shift felt inside the child's mind?"
Variation C · the hypothetical student
0:21
آپ نے 11:20 پر بچوں سے کہا "ہمیشہ رائٹ سے۔" 05:00 پر آپ نے انہیں خود ایڈیشن دریافت کرنے دیا تھا۔ اگر کوئی بچہ پوچھے کہ "میں نے تو خود گن کر کر لیا تھا، رول کیوں؟" — تو آپ کیسے بتائیں گی؟
"At 11:20 you told them 'always from the right.' At 05:00 you'd let them discover addition themselves. If a child asked, 'but I worked it out on my own — why the rule?' — how would you answer?"

All three cite the real timestamps from her lesson, surface the same teaching moment, and contain no advice and no judgement. But they read like three different thoughtful coaches. Variation C even invents a student's voice — something a fill-in-the-blank template categorically cannot do.

Better questions, for a fraction of the cost

It folds the lesson analysis and the questions into one cheaper, better model.

Cost to ask one teacher's questions: old GPT-4o chain $0.086, new DeepSeek V3.2 $0.003 — 96% cheaper; full observation cost drops 24–32%

How the AI finds the moment

It reads the whole transcript looking for a contradiction — two moments where the teaching changed. That tension becomes the question.

Timeline — 05:00 children count on their fingers; 11:20 a rule arrives: always start from the unit
8 guardrails every question must pass; 5 frameworks supported including MEWAKA; ~4,200 reflective questions per month; ≤65 word ceiling

What happens after the recording

1
Read the whole lesson. One pass pulls out the through-line, the key moments, and the moments a child's answer revealed their thinking.
2
Notice (Q1). Quote one real moment back, name the child if we can verify it, and place her inside it.
3
Listen, then adapt (Q2). The next question reacts to what she actually said — not a script — moving to a different moment on the same thread.
4
Commit (Q3). It closes on her own first thought for the next time she meets that specific kind of moment — a plan, in her words.

The rules a question must survive

Eight checks run automatically after each question is written. Fail one, and it's rewritten — up to three times — before a teacher ever sees it.

01
Stays under the cognitive-load ceiling — no rambling.
02
Cannot open with a yes/no stem (did · was · can · کیا) — open inquiry only.
03
No advice wearing a question mark — no should, try, chahiye.
04
Anchored to real evidence — a timestamp or the teacher's own words.
05
References what was done, never who the teacher is.
06
The closing question names a cue, a moment, and an observable action.
07·08
Genuinely new — checked against her last sessions by fingerprint and meaning.

The question is a voice note now — and it speaks Arabic without being taught.

Reflective questions arrive as native WhatsApp voice notes, in the teacher's own language. And because the language layer runs on principles rather than hard-coded rules, when we pointed it at a real Arabic lesson with no Arabic-specific code, it wrote natural, gender-neutral Arabic — switching to English for the subject terms — on the first try.

LIVE in production · May 2026