Online Python and AI classes in Huaraz, Peru

INEI sources: demographic indicators of the Censos Nacionales 2025, table 1 (reference day 15 September 2025); the 2017 sociodemographic profile, table 1.9, for the city; and the 2017 directory of populated places for Áncash, listing every settlement and how high it sits. The settlement rows are pieces of the city row, so adding them would count people twice.

From INEI's household survey (ENAHO) as printed in the Compendio Estadístico Perú 2026, volume 1, tables 11.5, 11.6 and 11.13. Ten years earlier, in 2015, only 16.2 per cent of Áncash homes had an internet service.

This is our usual pattern for a group class of fifteen to twenty learners of similar age. One-to-one classes follow the same shape with more time on the learner's own questions.

Two things make it work online. First, the learner always has the keyboard; the mentor never types the solution for them, but asks the next question until they find it. Second, the hour is fixed, so learning becomes a habit rather than something squeezed in when there is time. If you would like more on what the courses cover, our Python course page for Peru sets out the syllabus in detail.

1. Children, 9 to 12: AI from the inside. Children write tiny programs that learn from examples, such as a guesser that improves with each answer, and see it get things wrong. No chatbot accounts; the point is understanding, not using a product.
2. Teenagers, 13 to 18: models and honest testing. Teenagers train simple models on real data, then test them on data the model has not seen. They may use AI coding tools only where the tool's own age rules allow; some need a parent's permission until 18 and some, Claude among them, are for adults only.
3. University students and adults: AI as a colleague to check. Generative AI, prompting, automation and agents, with one rule on top: if the AI produced it, a person reads, tests and explains it before it is used.

At every age we teach thinking over copying. Paste in something you cannot explain and the mentor will have you reconstruct it piece by piece; those ten minutes tend to teach more than the rest of the week. The glacier project below is a good example: the first answer is easy to compute and wrong in an interesting way, and only reasoning about the data shows why.

import csv
import statistics

shortcut, measured = [], []
with open("llaca-inaigem.csv", encoding="utf-8-sig") as f:
    for row in csv.DictReader(f, delimiter=";"):
        low = float(row["TEMPERATURA_MINIMA"])
        high = float(row["TEMPERATURA_MAXIMA"])
        shortcut.append((low + high) / 2)
        measured.append(float(row["TEMPERATURA_PROMEDIO"]))

print("Days:", len(measured))
print("Shortcut average:", round(statistics.mean(shortcut), 2))
print("Station average:", round(statistics.mean(measured), 2))

The shortcut comes out half a degree warmer. On a glacier where the average day sits under three degrees, half a degree is a lot. But an average of averages can hide what is happening day by day, so Version 2 compares the two on every single day, counts how often the shortcut is too warm, and finds the days where it misses most.

import csv

warmer, gaps = 0, []
with open("llaca-inaigem.csv", encoding="utf-8-sig") as f:
    rows = list(csv.DictReader(f, delimiter=";"))

for row in rows:
    low = float(row["TEMPERATURA_MINIMA"])
    high = float(row["TEMPERATURA_MAXIMA"])
    mean = float(row["TEMPERATURA_PROMEDIO"])
    gap = (low + high) / 2 - mean
    if gap > 0:
        warmer += 1
    gaps.append((round(gap, 2), row["FECHA_REGISTRO"], low, mean, high))

gaps.sort()
print("Days the shortcut runs warm:", warmer, "of", len(rows))
print("Largest gap:", gaps[-1])
print("Smallest gap:", gaps[0])

humid = [(float(r["HUMEDAD_RELATIVA_MINIMA"]) + float(r["HUMEDAD_RELATIVA_MAXIMA"])) / 2
         - float(r["HUMEDAD_RELATIVA_PROMEDIO"]) for r in rows]
print("Humidity shortcut, average gap:", round(sum(humid) / len(humid), 2))

The shortcut is too warm on 140 of the 147 days, so this is a pattern, not bad luck. The worst day, 12 February 2025, had a minimum of 0.37 and a maximum of 6.26, so the shortcut says 3.32, while the station's own average was 1.87. Our reading of the shape: the warmest moments are short, perhaps when the sun breaks through, while most of the day stays near the cold end, so the middle of the two extremes sits above the real average. For humidity the bias flips: the shortcut runs almost three points too dry, which on the same reading would mean the air stays close to saturation most of the day and dries out only briefly.

Data: INAIGEM, Llaca glacier weather station (EMA-LLACA-1), datosabiertos.gob.pe, licence ODC-By. Monthly figures, gaps and counts are our own. The coldest reading in the file is -1.71 °C on 15 January 2025 and the warmest 8.94 °C on 10 January 2025; the minimum fell below zero on 13 of the 147 days. December starts on the 2nd and April ends on the 27th, as the file does.

1. Children, 9 to 12. Draw a day as a line of 24 dots, most of them cold and three of them warm. Work out the middle of the coldest and warmest dot, then the true average of all 24. Which is bigger?
2. Teenagers, 13 to 15. Run both versions and explain in four sentences why the shortcut runs warm for temperature and dry for humidity.
3. Teenagers, 15 to 18. Group the gaps by month. Is the shortcut worse in the wettest months? What would you need from INAIGEM to test the sunshine explanation?
4. University students and adults. Fit a simple correction, shortcut minus a constant, using December to February, and test it on March and April. How much of the error does it remove, and is it honest to use it on another station?

Peru stays five hours behind UTC through the year and does not change its clocks (tz database zone America/Lima; INACAL keeps the official hour). India stays five hours thirty minutes ahead of UTC, so it is always 10 hours 30 minutes ahead of Huaraz.

Students at the university often prefer an early-morning class before lectures or a late one after; secundaria students take theirs in the late afternoon; adults whose working weeks change from month to month often move to one-to-one classes at irregular hours. Booking happens in Huaraz time, and with no clock changes in either country the slot never drifts. For the wider picture, see coding and AI classes across Peru.