跪拜 Guibai
← Back to the summary

The 10 AI Task Types That Actually Cut Dev Time—and the 3 to Start With

Insert image description here

By now, we have learned quite a few AI programming methods:

This article does not introduce new concepts, but instead compiles a checklist worth bookmarking:

In day-to-day development, exactly what kinds of time can AI save us?

1. The 10 types of tasks AI programming is best suited to help with

Task Type What AI can do for you What to watch out for when using it
1. Requirement sorting Restate requirements, supplement questions, split tasks Final business rules must be confirmed manually
2. Technical plan Outline flows, data structures, and file changes Don't blindly adopt complex architectures
3. Code explanation Explain unfamiliar syntax, functions, and call chains After understanding, run and verify yourself
4. Basic code generation Generate drafts of functions, components, and interfaces Don't copy directly to production
5. Error troubleshooting Explain errors, organize troubleshooting steps Provide the full error and context
6. Unit tests List test scenarios, generate test code Check whether assertions are actually effective
7. Code optimization Improve readability, discover edge cases Performance optimization needs data backing
8. Documentation writing Generate READMEs, API docs, and changelogs Must not fabricate project facts
9. Repetitive work Process CSV, logs, files, and batch data Trial-run on a small amount of data first
10. Code review Check risks, omissions, and scope of changes AI cannot replace formal review

These 10 types of tasks share one common trait:

AI can help you organize and accelerate first, but the final result still needs testing and judgment.

2. The 3 types of tasks most worth prioritizing

If you haven't yet formed a daily usage habit, you don't need to try everything at once.

It's recommended to start with the following 3 types.

1. Let AI help you read code

When encountering unfamiliar code, let AI explain it first rather than modifying it directly.

Please explain the following JavaScript code.

Requirements:
1. First explain the overall functionality.
2. Explain the key steps in execution order.
3. Explain the role of each variable and function.
4. Give a simple input and output example.
5. Point out what I need to pay special attention to.

Code:
[Paste code]

This type of task suits learning new projects, reading legacy code, and understanding error contexts.

2. Let AI help you troubleshoot errors

Don't just send a message saying "the code is erroring out."

You should supplement with complete information:

Please help me analyze the following code error.

Runtime environment:
- Language/framework: [Fill in]
- Version: [Fill in]

Reproduction steps:
1. [Step 1]
2. [Step 2]
3. [The operation that triggers the error]

Full error:
[Paste the complete error and call stack]

Relevant code:
[Paste the code near where the error occurs]

Expected result:
[What should have happened]

Actual result:
[What actually happened]

First explain the most likely cause,
then give troubleshooting steps ordered by priority.
Don't rewrite the entire file at the start.

The more complete the context, the easier it is for AI to give targeted suggestions.

3. Let AI help you generate test scenarios

Often the problem isn't not knowing how to write tests, but not knowing what should be tested.

You can first let AI list scenarios:

Based on the following function and business rules,
list normal, boundary, and exception test scenarios.

Requirements:
1. Don't write test code yet.
2. Each scenario includes input, expected result, and test purpose.
3. Especially check null values, type errors, and max/min values.

Business rules:
[Paste rules]

Function:
[Paste code]

After confirming the test scenarios, then let AI generate code using the project's existing test framework.

3. When to use the other 7 types of tasks

1. Requirement sorting

Suitable when you've just received requirements and the description is relatively vague.

You can let AI help you distinguish:

2. Technical plan

Suitable before you start coding.

You can let AI output:

Confirm the plan first, then let AI write code; it's usually easier to control the scope of changes.

3. Basic code generation

Suitable for generating small functions with clear boundaries:

The smaller the task, the easier it is to understand and verify.

4. Code optimization

Suitable for letting AI first review:

Don't just say "help me optimize"; also specify whether the goal is readability, correctness, or performance.

5. Documentation writing

Suitable for organizing:

It's best to give AI the real code, configuration, and response data, letting it organize based on facts; don't let it guess.

6. Repetitive work

AI is very suitable for helping generate one-off scripts, for example:

Before running such scripts, back up data first and verify results with a small sample.

7. Code review

Before submitting code, you can let AI first check:

But AI review is only an auxiliary check; it cannot replace team review, automated testing, and security audits.

4. An AI programming workflow suitable for daily use

You don't need to use complex prompts for every task.

You can remember this simple workflow:

Clarify the problem to solve today
  ↓
Supplement project background and constraints
  ↓
Let AI analyze or propose a plan first
  ↓
Complete only one small step
  ↓
Run, test, and confirm
  ↓
Record the final result

For example, today you need to fix an error:

Problem: Error when clicking the save button.
Background: Vue 3 project, using Node.js 20.
Goal: Restore save functionality.
Constraints: Don't modify the API, don't add new dependencies.
Verification: Test successful save, empty input, and API failure.

These few sentences are already much clearer than "help me fix the save function."

5. What not to fully hand over to AI

AI can save time, but the following content requires special caution:

You can judge with one sentence:

The higher the cost of failure, the more manual confirmation and thorough testing is needed.

6. A bookmarkable usage checklist

[ ] Did I let AI explain a piece of code today?
[ ] Did I let AI help sort out a requirement today?
[ ] Did I let AI propose an implementation plan first today?
[ ] Did I let AI help troubleshoot an error today?
[ ] Did I supplement a boundary test today?
[ ] Did I let AI check the scope of code changes today?
[ ] Did I organize a README or API documentation today?
[ ] Did I use AI to handle a repetitive task today?
[ ] Did I verify the results generated by AI today?
[ ] Did I record a reusable prompt today?

You don't need to complete all 10 items every day.

Just pick one item, complete the loop of "ask, execute, verify," and you are already accumulating your own AI programming experience.

Summary

The tasks AI is best suited to help developers save time on are mainly concentrated in:

But improving efficiency does not mean reducing judgment.

The truly effective way to use it is:

Let AI handle repetitive work
Let AI provide ideas and drafts
Take responsibility for judgment, verification, and delivery yourself

You can condense this week's content into one sentence:

AI saves you execution time, while you must retain the ability to understand and judge.

In the next article, we will enter a very important topic:

"Privacy and Security in AI Programming: What Information Cannot Be Submitted Directly"

✍Insist on original content, seeking follows, likes, and bookmarks