Build with AI without copying: live classes for Peru
An AI assistant can write a working function faster than any student. It can also write one that works on the example you gave it and breaks on the next. The companies that make these tools say so in their own documents. This page shows what they say, which tools a teenager in Peru may use, and the routine we teach so that what you build is still yours.
Live classes since 2020 · 10,000+ students worldwide · AI output treated as a draft, never as the answer
The short answer
Building with AI without copying means you let an assistant draft code but accept nothing you cannot read, predict, test and explain yourself. Modern Age Coders teaches this live and online, in English, to teenagers from 13, with a parent's permission where a tool requires it, and for university students and adults in Peru. Mentors in India teach a fixed review routine, starting from the AI makers' own warning that output can be wrong. You choose a group of fifteen to twenty or one-to-one lessons, at flexible hours.
Start here
Three courses where AI is a reviewed collaborator
One for teenagers, one for university students and one for working adults. In all three the learner plans and tests; the assistant drafts.
What do AI companies say about checking AI-generated code?
They say check all of it. Each major assistant's own documentation or terms tells users that output can be wrong while looking right, and that reviewing it is the user's job. These are not critics speaking; they are the makers.
What three AI makers tell their own users about output, quoted from their documents
Maker and document
What it says
What that means for a learner
GitHub, application card for Copilot inline suggestions
"You should always review Copilot's suggestions before accepting them"
Accepting a suggestion is a decision you make, so you need to be able to make it
GitHub, same document
"Hallucinations are a known risk of large language models and are a key reason that human review of AI-generated output is important."
Code can call functions that do not exist or misuse ones that do
Anthropic, Consumer Terms of Service (effective 8 October 2025)
"Outputs may not always be accurate and may contain material inaccuracies even if they appear accurate because of their level of detail or specificity."
A detailed, confident answer is not evidence that it is right
Anthropic, same terms
"You should not rely on any Outputs or Actions without independently confirming their accuracy."
Confirming means a test or a check that does not come from the same tool
Google, Gemini Apps Privacy Hub (updated 26 October 2026)
"Gemini can make mistakes." and LLM experiences "can hallucinate and present inaccurate information as factual."
The same caution applies whichever assistant you use
Sources: GitHub Docs, "Application card: GitHub Copilot inline suggestions"; Anthropic, Consumer Terms of Service; Google, Gemini Apps Privacy Hub. All read in English on 9 November 2026. OpenAI's site refused our request that day, so ChatGPT appears below only through its age rule, quoted from the vendor page in our July 2026 audit.
Why a learner should take this seriously
A professional who reviews AI code is comparing it with years of their own. A learner has nothing to compare it with yet, which makes copying more tempting and more risky at the same time. The answer is not to avoid the tools. It is to build the judgement first, then use the tools where that judgement can catch their mistakes.
In practice that means a learner should be able to write a small version of the thing on their own before asking an assistant for the big version.
What "independently confirming" looks like
Anthropic's phrase is the useful one. Confirming independently means a check that does not depend on the tool's own explanation: running the code on inputs you chose, tracing it by hand, reading the official documentation for a function, or comparing with a version you wrote.
Asking the same assistant "are you sure?" is not independent. Its second answer comes from the same place as its first.
Age rules
Which AI coding tools can a teenager in Peru use?
It depends on the tool, and the rule is in each maker's own terms, not in a school policy. Some allow 13-year-olds with a parent's permission; some require users to be adults. We follow the strictest reading in class.
What the terms say
Anthropic's consumer terms say: "You must be at least 18 years old or the minimum age required to consent to use the Services in your location, whichever is higher." So Claude is an adults-only tool in our classes.
OpenAI's terms, as quoted in our audit of July 2026, say that users under 18 "must have your parent or legal guardian's permission". Microsoft says users must be at least 13 to use Copilot. Google's Gemini help says a Family Link account cannot use the Gemini web app, so a supervised child account cannot use it there.
What we do with that
Under 13: no chatbots in class. Children learn how AI works through projects that do not need an account, such as training a tiny model on their own examples.
From 13 to 17: tools whose terms allow the age, with the parent's permission where the terms ask for it, and only after the learner has written their own first version. From 18: every tool the course needs, including coding agents. The full list, tool by tool, is on our AI tools age guide.
Tool terms change. We re-read them before a course uses a tool, and a parent can ask the mentor at any time which tools their teenager is using and under whose account. For the AI side of the curriculum for secundaria, our Spanish page on IA para adolescentes explains what we teach about how models work.
A real lesson
How can AI code pass a test and still be wrong?
By being tested on the one case it happens to handle. The lesson below is ours. The first function is the kind of short, tidy answer an assistant often gives to "remove the blank names from a list"; we wrote it ourselves to make the point, and we never show students a real tool's output as if it were a fixed fact.
Version 1: the tidy answerPython
# A plausible answer to: "remove the blank names from a list"
def drop_blanks(names):
for name in names:
if name == '':
names.remove(name)
return names
print(drop_blanks(['Ana', '', 'Luis', '', 'Rosa']))
print(drop_blanks(['Ana', '', '', 'Luis', 'Rosa']))
print(drop_blanks(['', '', '', 'Rosa']))
It reads well: loop over the names, remove each blank one, return the list. Before running it, the mentor asks the learner to predict all three printed lines.
Copied verbatim after we ran it under Python 3. The first test passes. The second leaves a blank behind. The third leaves two. A learner who tried only the first list would have shipped it.
The pattern in the failures is the clue: the function fails when two blanks sit next to each other. Rather than ask the assistant to "fix it", the learner makes the function show its work, printing what the loop sees at each step and what the list looks like at that moment.
Version 2: make the loop show what it seesPython
def drop_blanks_traced(names):
for i, name in enumerate(names):
print('step', i, 'sees', repr(name), 'list is now', names)
if name == '':
names.remove(name)
return names
print(drop_blanks_traced(['Ana', '', '', 'Luis', 'Rosa']))
enumerate gives the step number alongside each name, and repr prints an empty string as two quote marks, so it is visible.
What Version 2 printsOutput
step 0 sees 'Ana' list is now ['Ana', '', '', 'Luis', 'Rosa']
step 1 sees '' list is now ['Ana', '', '', 'Luis', 'Rosa']
step 2 sees 'Luis' list is now ['Ana', '', 'Luis', 'Rosa']
step 3 sees 'Rosa' list is now ['Ana', '', 'Luis', 'Rosa']
['Ana', '', 'Luis', 'Rosa']
Exact output from our own run. At step 1 the loop sees the first blank and removes it. Every later name slides one place to the left, so the second blank moves into position 1, which the loop has already visited. At step 2 it sees Luis. The second blank is never looked at.
Name the cause in one sentence. "Removing an item shifts the rest left, but the loop still moves forward one place, so it skips whatever slid into the gap." If the learner can say that, they understand it.
Check the official source. The Python tutorial warns about this exact pattern: "Code that modifies a collection while iterating over that same collection can be tricky to get right." It suggests looping over a copy or building a new collection instead.
Prefer a new list. Building a fresh list leaves the original alone, which also avoids a second, quieter problem: Version 1 changes the list the caller passed in.
Write the tests before the fix. Adjacent blanks, blanks at the start, a list of only blanks and an empty list, plus a check that the input is untouched.
Version 3: build a new list, then test hardPython
def drop_blanks(names):
return [name for name in names if name != '']
tests = [
['Ana', '', 'Luis', '', 'Rosa'],
['Ana', '', '', 'Luis', 'Rosa'],
['', '', '', 'Rosa'],
['', ''],
[],
]
for t in tests:
original = list(t)
result = drop_blanks(t)
print(result, '| no blanks left:', '' not in result, '| input untouched:', t == original)
The comprehension reads as one English sentence: keep each name that is not blank. list(t) takes a copy before the call, so the last check can prove the input was not modified.
Exact output from our own run: five cases, no blanks left in any result, and every input list exactly as it was before the call.
Could an assistant have produced Version 3 first time? Often, yes. The point of the lesson is that the learner can now tell Version 1 from Version 3 without running them, and knows which four tests to write next time. That is the difference between building with AI and copying from it.
Our review routine
How do you review AI-generated code before using it?
With the same five steps every time, in the same order. The routine is ours and it is deliberately boring: a habit only works if it does not depend on how confident you feel about a particular answer.
State the job and the tests first. Before prompting, write in plain English what the code must do and three to five cases with their expected results, including at least one awkward one.
Read it aloud, line by line. For each line, say what it does. Mark any line you cannot explain; marked lines are not accepted until you can.
Predict, then run your tests. Say what each test will print before running. A wrong prediction on a passing test still means you do not understand something.
Shrink it. If the answer is longer than the problem deserves, rewrite the part you understand in fewer lines, or ask for a smaller version. Long code hides bugs from beginners.
Explain the change you would make. Pick one new requirement, such as "also remove names that are only spaces", and describe where it goes. If you cannot, you are not ready to own this code.
For a teenager
Steps 1 to 3 every time, step 4 when the answer is longer than a screen. The mentor asks for step 5 aloud at the end of each project.
For a university student
All five steps, plus version control: every accepted change is its own commit with a message that says why, so a bad suggestion can be undone.
For a professional
All five, plus a check against the project's own rules on data and security. Agents that edit many files get their changes reviewed as if a new colleague had written them.
For the wider picture, including what vibe coding is and where it differs from this routine, our Spanish guide vibe coding en inglés is the place to start. Parents of younger children who would rather wait on AI tools can see what we teach before any assistant appears on our Spanish page about clases de programación para adolescentes.
AI in Peruvian education
How is AI being used in Peruvian schools and competitions?
Cautiously and in specific places. Peruvian institutions have begun using AI for planning and running AI projects, while the computing olympiad keeps it out of the contest room entirely.
Planning in rural secondary schools
MINEDU announced in December 2025 that from the following year it would use AI for learning planning in 205 rural secondary schools serving about 27,000 students. The tool supports teachers; it does not do the students' work.
Student projects
MINEDU's Hackathon 2025, its fourth edition, rewarded teams of students and teachers for projects in AI and robotics, and reported that participation by girls grew 28% between 2023 and 2025.
The contest room
UTEC's rules for the 2025 Peruvian Computing Olympiad ban generative AI for solving problems. A teenager who aims at the OPI needs to code alone, which is one more reason to build skill before leaning on tools.
We are not connected with MINEDU, UTEC or any school, and our classes do not count towards Peruvian grades. They run beside school and work, and a learner who finishes a course receives our own certificate of completion.
Taught in English
Are the classes taught in Spanish?
No. Lessons are live, online and given only in English. The mentors, who work from India, keep their speech clear and unhurried and follow the pace and accent of the person in front of them. For AI work, English carries a practical advantage.
Prompts, errors and documentation
Most of the documentation an AI assistant was trained on, and most of the error messages it will explain, are in English. A learner who can write a precise prompt in English, and read the documentation that confirms or contradicts the answer, can check the tool against its sources.
In class, learners write their task statements and test descriptions in English, which doubles as practice in saying exactly what they mean.
If your English is still growing
That is normal and fine. Mentors slow down, repeat in simpler words and type key phrases into the chat. Many adults in Peru read English far better than they speak it, and the classes give daily speaking practice in a low-pressure setting.
You may write to us on WhatsApp in Spanish whenever you like; the lessons themselves stay in English.
For Lima learners at bilingual or international schools, our page on English-medium coding classes in Lima describes how classes fit beside an English-language timetable.
Peru time
When can I take a live class from Peru?
Whenever suits you. Mentors work from India, and the gap between the two clocks never changes, because neither country moves its clocks during the year.
Peru time beside India time (conversions only; you choose the hour)
Peru time
India time
Day in India
6:30 a.m.
5:00 p.m.
Same day
8:30 a.m.
7:00 p.m.
Same day
11:30 a.m.
10:00 p.m.
Same day
2:30 p.m.
1:00 a.m.
Next day
6:30 p.m.
5:00 a.m.
Next day
9:00 p.m.
7:30 a.m.
Next day
Peru is on UTC minus 5 throughout the year and does not use daylight saving (tz database, America/Lima). India is on UTC plus 5:30, also without daylight saving, so India is always 10 hours 30 minutes ahead of Peru.
Working adults in Lima often take an early morning class before the office; university students tend to pick late evenings. Either way, you book in Peru time and the conversion is our job.
Progress
How does judgement with AI tools grow?
Five levels of our own. Most learners start at 1 or 2, whatever their age; what moves them up is practice with the review routine, not more prompting.
Five levels of judgement when coding with an AI assistant (ours)
Level
Typical behaviour
Next step
1. Paste and hope
Copies output, runs it once, moves on if no error appears
Write the expected result before running
2. Test the example
Checks the case given in the prompt only
Add an awkward case: empty, repeated, reversed
3. Read every line
Can explain each line and marks the ones they cannot
Trace the marked lines by hand
4. Catch and fix
Finds a hidden failure and repairs it without asking the tool
Write the tests before generating anything
5. Direct the tool
Specifies the job and tests, judges drafts, rejects weak ones
Review a classmate's AI-assisted code
For parents
Ask your teenager which level they are on this week and what moved them. A learner at level 3 or above can tell you about a time the AI was wrong.
If the honest answer is level 1, that is where most beginners start. The routine is designed to move them.
For learners
Keep a short log of every time an assistant gave you something wrong and how you caught it. It becomes a list of your own blind spots.
The fastest route to level 5 is writing the tests before the prompt.
More courses
Nine more courses for building with AI
All live, in English, with one mentor, and all using AI tools within each maker's age rules.
I
Teenagers
Ages 13 to 17
BWA / TEEN / 01
Python for teens
The foundation that lets a teenager judge generated Python.
A weather forecasting site with live conditions for any location.
by Krish
Fees
Fees for learners in Peru
Plans are priced monthly in US dollars and charged to a debit or credit card. Your bank does the conversion to soles at its own rate, which is why the amount you see on your statement may differ by a little each month.
What families and learners wrote about us on Google, unedited.
★★★★★
"The one step solution for my son. Modern Age Coders make learning coding so simple that kids love it. The teachers explain complex concepts clearly with practical exercises and interactive content."
Ria Mukherjee
Parent
★★★★★
"Modern Age Coders has been a game-changer for me. I struggled to grasp IT concepts and coding before joining, but their classes transformed everything. I can now confidently write complex programs with ease."
Samriddha Mondal
Student
★★★★★
"One of the most wonderful education centres out there. Education is not limited to school syllabus but focuses on skill development."
Vansh Agarwal
Student
★★★★★
"My child Dhairya is really enjoying the Modern Age Coders classes. This is his first online class and he eagerly looks forward to it. I can already see his improvement, and the teachers are very cooperative."
Sonam Oswal
Parent of Dhairya
★★★★★
"Modern Age Coders have wonderful teachers who teach in a clear, easy and practical way. The teacher boosts students' confidence and inspires them to learn without hesitation."
Sonu Goyal
Parent
★★★★★
"I highly recommend this computer coding class! The teachers are incredibly knowledgeable and passionate about coding."
Ritu Kedia
Parent
Learners in Peru ask
Building with AI without copying: questions and answers
Are the classes in Spanish?
No. Every class is live, online and taught only in English. Mentors based in India keep their English plain and steady and adapt to how quickly you follow. You can write to us in Spanish on WhatsApp, but lessons stay in English.
Is using AI to code cheating?
Not in itself. It becomes copying when you accept code you cannot explain or test. In a contest that bans AI, such as the OPI, or in homework where the teacher forbids it, using it anyway is cheating, and we teach learners to follow those rules.
What age can my child start using AI coding tools?
In our classes, from 13, with tools whose own terms allow it and a parent's permission where the terms ask for it. Claude requires users to be 18, so only adults use it. Children under 13 learn how AI works without chatbot accounts.
Do I need to know how to code before using AI?
You need enough to read and test what the tool writes. Complete beginners spend their first weeks writing small programs alone, then bring in an assistant once they can check its work.
Which tools do you use?
It depends on the course and the learner's age: coding assistants in an editor, chat assistants for explanations, and for adults, coding agents that edit whole projects. Tool terms are re-read before a course uses them.
What time are the classes?
Any day, any time, flexible: choose a batch, or for a very personalised plan choose 1-on-1 at the hour that suits your family. India is 10 hours 30 minutes ahead of Peru all year, so 8:30 a.m. in Lima is 7:00 p.m. in India.
Can a professional with a full-time job join?
Yes. Many adults take one-to-one classes early in the morning or late in the evening, Peru time, and bring a real problem from work to build during the course.
Will I get a certificate?
You receive our own certificate of completion when you finish a course. It is not a university credit or an official Peruvian qualification, and we are not connected with MINEDU or any university.
What if the AI gives a different answer each time?
That is normal for these tools and one more reason the tests come first. If two different answers both pass your tests, your tests are probably too easy, and the mentor will help you write a harder one.
How do I pay?
By card, monthly, in US dollars; your bank converts the charge into soles. The first class is free and needs no card.
Elsewhere on this site
More on AI and code
English pages that go further into the tools and the thinking behind them.
In the free class you try one short task twice: once alone and once with an assistant, then you review the assistant's version with the mentor. It shows quickly where your judgement already is, and the mentor suggests a course, a level and the tools that fit your age.
Want to look around first? How we teach describes a live class, and the course list has every syllabus.
Live online classes with mentors in India, taught in English only. WhatsApp is the quickest way to reach us, in Spanish or English.
Want a class sooner? Book a priority demo
A priority demo is a full class of about 45 to 60 minutes. After the class the mentor sends a written skill report and a personal learning roadmap on WhatsApp. If you enrol, the demo fee is adjusted against your first month's fee. Share your demo Order ID with us on WhatsApp and we take it off.
