DEV Community: Faizan Altaf

How the rupee works: what sets the dollar rate in Pakistan

Faizan Altaf — Tue, 03 Nov 2026 07:21:24 +0000

Every morning in Pakistan, someone checks the dollar rate. Shopkeepers, importers, families waiting on money from abroad, students paying foreign university fees — the number matters to all of them. But few people know what actually sets it.

This is the plain-English version: how the rupee's price is decided, who decides it, and what moves it.

The rupee has a price, like anything else

The exchange rate is simply the price of one currency in terms of another. When the rate is Rs276.75 to the dollar, it means you need Rs276.75 to buy one US dollar. Like any price, it moves with supply and demand: when more people want dollars than rupees, the dollar gets more expensive and the rupee falls. When dollars flow in, the rupee steadies.

Most of that buying and selling happens in the interbank market, where banks trade dollars with each other and with the State Bank of Pakistan. The rate you see on the news is usually the interbank closing rate — the rate at the end of the trading day. The rate at your local exchange company is close to it, but not identical: exchange companies add their own margin, and cash dollars can cost a little more than bank transfers.

Who actually sets the rate?

Nobody sets the rate by decree. Pakistan has a market-based exchange rate, which means the rate moves with trading. But the State Bank is not a bystander. It watches the market every day, and it can buy or sell dollars to smooth sharp swings — selling dollars when the rupee is falling fast, buying them when it wants to build reserves. It does this quietly, and it does not announce every intervention.

The State Bank publishes the official closing rate each trading day. That is the number that counts for banks, businesses and the news. On the last published close before this piece was written, the rate was Rs276.75 to the dollar. The live market rate on the morning of Monday 2 November 2026 was trading a little higher, around Rs277.44 — a reminder that the published close is a snapshot of yesterday, and the market keeps moving.

What pushes the rupee down

Four things matter most.

Imports. Pakistan buys far more than it sells. It imports oil, gas, machinery, food and medicine, and pays for them in dollars. In the first three months of the current financial year, the gap between what Pakistan bought and sold — the current account deficit — was $874 million, according to State Bank balance-of-payments data. Every dollar of that gap is demand for foreign currency, and it pushes the rupee down.

Debt payments. Pakistan owes money abroad, and those loans are repaid in dollars. When a big repayment falls due, the demand for dollars jumps.

Confidence. If people expect the rupee to fall, they buy dollars now — importers pay early, savers convert rupees, businesses hold dollars. That expectation alone can move the rate. It is why rumours and political uncertainty show up in the exchange rate so quickly.

The street rate. If the gap between the official rate and the rate on the informal market widens, some money moves off the books. That weakens the official market and makes the published figures look worse than the money actually moving.

What holds the rupee up

The strongest support comes from overseas Pakistanis. They sent home a record $41.6 billion in the 2025–26 financial year, according to the State Bank — more than Pakistan earned from everything it sold abroad. The first three months of the new financial year brought in $10.87 billion. That flow supplies the dollars Pakistan needs to pay its import bills, supports the central bank's foreign exchange reserves, and goes straight into family budgets. I explained the full picture — where the money comes from and why it matters — in this Findivoza piece: Pakistan's economy: remittances hit $10.9bn as imports outpace exports.

Two other supports matter. IT exports crossed $3 billion in the first ten months of 2025–26, and IT freelancers earned $959 million in the full year, up 49 per cent. And foreign loans and IMF support, when they arrive, add dollars to the central bank's reserves — which is why an IMF programme announcement often steadies the rupee before a single dollar is spent.

Why the rate matters to ordinary people

If your family receives money from abroad, the rate decides how many rupees arrive for each dollar sent. If you buy petrol, cooking oil or imported medicine, the rate is baked into the price — a weaker rupee means dearer imports, and that feeds straight into inflation. If you are a student paying fees abroad, or a business importing stock, the rate is a direct cost.

That is also why the rate is political. A falling rupee feels like the country getting poorer, because in rupee terms, everything imported costs more. A steady rupee feels like stability, even when the underlying pressures have not gone away.

What to watch

If you want to understand where the rupee is heading, watch five numbers: monthly remittances (about $3.5 billion a month on average in 2026 — if they hold, the rupee has a floor); the current account deficit (the monthly gap between imports and exports); the State Bank's reserves (how many dollars it holds to defend the rate); the gap between the official rate and the street rate (a widening gap means trouble); and oil prices (Pakistan imports almost all its oil, so a dearer barrel means more dollars going out).

None of these predicts the rate tomorrow. But together they tell you whether the pressure is building or easing — and that is more useful than any single morning's number.

The bottom line

The rupee's price is not set by one person or one decision. It is the result of millions of transactions: a nurse in Manchester sending money home, an importer in Karachi paying for oil, a student in Lahore paying a university in London, the State Bank buying or selling to steady the market. The rate you see each morning is the sum of all of that.

Understanding it does not make the rupee stronger. But it does mean the number on the screen stops being a mystery — and you can tell the difference between a normal wobble and a real warning sign.

Faizan Altaf is the author of Findivoza, a plain-English news blog explaining Pakistan and the world simply.

Sources

State Bank of Pakistan, workers' remittances data (FY2025–26: $41.6bn; July–September 2026: $10.87bn); balance of payments data (current account deficit $874m, Q1 FY27); IT freelancer earnings $959m, +49% (FY2025–26).
State Bank of Pakistan, interbank closing rate (last published close Rs276.75 per US dollar, before 2 Nov 2026); live market rate ~Rs277.44 on the morning of 2 Nov 2026 (market rate, indicative only).

How AI search answers work — and why they sometimes invent facts

Faizan Altaf — Mon, 02 Nov 2026 10:03:02 +0000

You ask an AI search tool a question. It answers in a second, in clean paragraphs, with confidence. Sometimes the answer is excellent. Sometimes it invents a date, a name or a reference that never existed — and says it with the same confidence.

If you build with these tools, or you just use them every day, it helps to know what is actually happening inside. This is the plain-English version, with enough of the mechanics to be useful.

It predicts words, it does not look things up

A large language model does not "know" facts the way a database does. It was trained on a huge amount of text, and what it learned is a statistical pattern: given these words, what words are most likely to come next?

When you ask a question, the model generates the most likely continuation, one token at a time. A token is roughly a word or part of a word. Each step, it scores every possible next token and picks one — usually the most likely, sometimes a slightly random one to keep the text natural.

That is why the output reads so fluently. It is also why the output can be wrong. The model is optimising for plausible text, not for truth. A made-up citation looks exactly like a real one to a system that only cares about what words usually follow other words.

AI search adds retrieval — but the model still writes the answer

Modern AI search tools bolt a second system onto the model: retrieval. The flow looks roughly like this:

Your question is turned into a search query.
The tool fetches real web pages — the same kind of results a search engine returns.
Those pages are chopped into passages and handed to the model as context.
The model writes an answer based on that context.

This is called retrieval-augmented generation, or RAG. It helps a lot: when the answer is in the retrieved passages, the model usually gets it right and can even cite the source. But notice step 4. The model is still writing. If the retrieved passages are thin, contradictory or missing the key fact, the model fills the gap the only way it knows how — by generating plausible text. That is where hallucinations come from in AI search: not from nowhere, but from the gap between what was retrieved and what the question needs.

Three things make that gap bigger: questions about very recent events (the search index may not have the latest page yet), questions with no good source on the web (the model has nothing solid to stand on), and questions that mix two facts (the model stitches together passages that were never meant to be combined).

Why dates, numbers and names fail first

If you watch AI answers fail, the failures cluster. Dates, numbers, names and references go wrong far more often than general explanations. There is a mechanical reason: these are the tokens with the least redundancy. In a sentence like "the meeting happened on 14 March", the date carries almost all the information, and there is no surrounding context that lets the model recover it if it guesses wrong. A general explanation has slack — many wordings are acceptable. A date has exactly one right answer, and the model has no way to check it unless the retrieved text contains it.

That is also why a confident tone means nothing. Confidence in the output is a property of the text generation, not a measurement of accuracy. The model does not know when it is guessing.

What actually reduces hallucinations

If you are building or evaluating these tools, four things matter most:

Better retrieval, not a bigger model. The single biggest improvement in AI search has come from fetching better passages — more sources, fresher pages, passages ranked by relevance to the exact question. A modest model with good retrieval beats a large model with poor retrieval on factual questions.

Grounding and citations. When the tool shows its sources and ties each claim to a passage, two things happen: the model is constrained to what it can support, and the user can check. A citation that does not actually support the claim is worse than no citation, because it borrows trust it has not earned.

Asking the model to say "I don't know". Models can be tuned to refuse when the retrieved evidence is weak. Most products resist this because users dislike refusals — but a refusal is more useful than a confident invention.

Verification passes. A second, cheaper check — does the answer's key claim appear in the retrieved text? — catches a large share of inventions before the user sees them.

None of these removes the problem entirely. They shrink it, and they make the remaining errors checkable.

How to use AI search without being misled

For everyday use, three habits cover most of the risk:

Treat numbers as unverified until you see the source. If an answer gives you a figure, click through to the source and check the figure is actually there. This is the single most valuable habit.
Ask for the source, then read it. "According to whom?" is the question that separates a good answer from a confident guess.
Never trust it for things that must be exact. Dates, legal details, medical doses, financial figures — verify those against a primary source, every time.

None of this means AI search is useless. For explaining a concept, summarising a long document you already have, or drafting a first version of something, it is genuinely good. It is a bad oracle and a good assistant — and the difference matters.

The same problem, in a classroom

This is not abstract. In Pakistan, teachers are adopting AI tools faster than any policy: the share of surveyed teachers who had used generative AI rose from 64 per cent in 2024 to about 93 per cent in 2026. Students use the same tools for homework. The most valuable skill in that setting is not using AI — it is checking it. Only about a quarter of the teachers surveyed get their students to critique what AI produces.

I wrote about what that looks like on the ground — the curriculum changes, the free subscriptions, the cheating problem and the access gap — in this explainer on Findivoza: How AI tools are changing Pakistani classrooms.

The lesson is the same whether you are a developer building a RAG pipeline or a student doing homework: the model writes plausible text. Your job is to check whether it is true.

Faizan Altaf is the author of Findivoza, a plain-English news blog explaining Pakistan and the world simply. He writes about technology, AI and the economy for readers who want the facts without the jargon.