Tutor vs ChatGPT for CS1010, CZ1003, and First-Year Programming

ChatGPT is the most-asked-about thing in my Telegram inbox right now. "Why pay for a tutor when ChatGPT is free?" It's a fair question and the honest answer is more nuanced than either side would tell you. AI tools are genuinely useful for some parts of learning to programme. They are also genuinely terrible for others. And because most first-year modules now have explicit policies on AI use, there's a third question: how do you use either tool honestly?

This post is for the student deciding where to spend time and money before semester starts.

Where ChatGPT genuinely beats a tutor

For these tasks, an AI is usually the right tool, often the only tool you need:

• Looking up syntax you forgot. "How do I read a file line by line in Python again?" gets a correct answer in 5 seconds.
• Explaining a single error message. Paste the traceback, get a plain-English explanation. A tutor would charge for the same thing and take longer.
• Generating boilerplate. A test scaffold, a CLI argument parser, a Flask hello-world. Use it.
• Translating between languages. "Show me the C version of this Python function" works well enough for learning purposes.
• Patient repetition. Asking the same beginner question fourteen different ways. A tutor will be polite about it; a model genuinely doesn't care.
• 2am availability. No scheduling, no minimum-billing, no awkwardness about the question being too basic.

If your need is any of the above, save the money and use the AI.

Where ChatGPT genuinely fails

Equally honestly, here are the failure modes I see in student code every week:

• Confidently wrong answers. The model will produce code that compiles, runs, and silently does the wrong thing. A tutor watching you trace the logic catches this in real time; an AI doesn't know it's wrong because it doesn't know what your assignment actually wants.
• Hallucinated module behaviour. Especially common with niche libraries (pandas chained-indexing edge cases, pytorch autograd quirks, JavaFX event handlers). The AI generates plausible-looking calls to functions that don't exist or behave differently from what it claims.
• Inability to read your assignment brief. Even if you paste the brief in, the model misses constraints that humans would catch: "must use recursion only", "must run in O(n log n)", "must not use external libraries".
• No mental model of your specific module. It doesn't know that CS1010 marks pointer arithmetic strictly, that CS2030S penalises imperative code in late assignments, that CZ1003 expects a specific commenting style. A tutor who has helped twenty students through the same module does.
• Building dependence instead of skill. The most common pattern I see: student uses AI to "help" with assignments 1 to 4, hits assignment 5 (which is a stepped-up version of the earlier ones), and now has nothing to fall back on because they never internalised what was happening.
• No accountability when it goes wrong. The model doesn't care if you fail. A human tutor whose Carousell listings are public has reputational skin in the game.

When a tutor earns the cost

Concrete scenarios where paying a human is the right call:

1. You've used AI on the early assignments and you can't actually defend the code. You're about to walk into a viva or a practical and you need someone to make sure you actually understand what's going on. AI can't do this; it's the exact gap a tutor fills.
2. The assignment requires you to invent something. Genuine project work, FYPs, capstones. The AI generates plausible code; the tutor helps you reason about whether the architecture makes sense for your specific scope.
3. You're stuck and stuck-ness is the bug. You've been spinning for three hours. A tutor can spot the false assumption you're operating under in five minutes; an AI will continue to politely confirm your false assumption while suggesting fixes that don't address the real cause.
4. You need someone who knows your marker. Singapore tertiary modules have very specific marking rubrics. CS2103T cares about Git discipline. CS2040S cares about complexity proofs. CZ1003 cares about clean output formatting. A tutor who has seen this rubric repeatedly knows what loses marks; the AI doesn't.
5. You want to learn, not just submit. This is the biggest one. If your goal is the grade, AI can probably get you a passing one. If your goal is to actually understand what you're writing, a tutor is the structural way to make that happen.

How to use AI honestly

Most Singapore tertiary institutions in 2026 have published policies on AI use. The two patterns I see:

• "Allowed with declaration", you can use AI but must declare what you used it for.
• "Allowed for learning, not for submission", you can use AI to study but must submit your own work.

Either policy makes the right behaviour pretty clear: use AI to understand, not to submit. The honest workflow:

1. Read the brief yourself first. Form your own first attempt.
2. When stuck, ask the AI to explain the concept, not give you the answer.
3. Write the code yourself, even if you copied the structure of an AI suggestion.
4. Test it yourself. Read every line and make sure you can explain it.
5. If you can't explain it, you don't understand it, and you shouldn't submit it.

If you're working with me, I'll often use the same workflow. I might use ChatGPT to look up an obscure pandas idiom in the middle of a session; I won't use it to write your assignment for you. The line is the same one your institution drew.

Cost comparison, since people always ask

• ChatGPT Free: S$0/month, with limits on the more capable models.
• ChatGPT Plus / Claude Pro / Gemini Advanced: S$20-30/month equivalent.
• Singapore freelance tutor: typically S$80-150/hour for an experienced working engineer (cheaper for student tutors, more for deep-specialist module experts). No monthly commitment.

Using both makes sense for most students: AI for the always-available basics, a tutor for the targeted work that needs a human. The combined monthly spend for a serious CS student during a tough week (one or two tutor sessions plus an AI subscription) is typically S$200-500. In quiet weeks it can drop to just the AI subscription.

Where I sit

I tutor Singapore students through CS1010, CZ1003, and the rest of the first-year syllabus. I won't pretend you should never use ChatGPT; I will help you use it well, defend your own code, and fill in the gaps where the AI is confidently wrong. If that fits, send a brief on Telegram or via the form and I'll be straight about whether the work suits a tutor or you can handle it with AI plus self-study.