How to use GenAI as an engineer

Day 7 of a hands-on, non-technical guide for executives

Start with the Introduction to series . This is Day 7

Today’s exercise makes sense only if you are somewhat technical (an engineer, a product manager, etc.). Still, I recommend you try this because it shows where white collar work in many domains is headed.

At the most basic level, you can ask questions and get answers. That’s like a chatbot. But there are higher levels that you should be aware of.

Image generated using Gemini

Level 1: Using Google Search

At the most basic level, you can simply query for code. For example, go to Google and search:

python code to get logprobs from openai

You can copy-paste this into your editor, but using a coding assistant goes so much more than that.

Level 2: Using Claude Code

At the next level, you can use code as an artifact and keep modifying it.

Go to https://claude.ai/ and type in the same query:

python code to get logprobs from openai

Now, have it edit the code in a rather simple way:

instead of 2.718, use math.e

Now, make it make a more meaningful change that involves complete redesign:

rewrite this code to work in conjunction with structured outputs. I want a single method that will take the input, the desired output structure as a Pydantic object, and return a Pydantic object along with the average logprob for each data element

Redesigned code in Claude

Level 3: Using within IDE

Level 2 is good for 0–1 use cases where you are building a new feature. But what if you want to update existing code? This is where AI assistants with IDE-integration like Cursor, Windsurf, Augment, GitHub Copilot, etc. start to shine.

Programmers can do things like pasting in an error message and asking the AI tool how they would fix the error. Or attaching a screenshot of their current UI and asking the AI what to modify to change the order in which different UI components appear.

Try this with Gemini CLI, since it can be helpful for interacting with all kinds of files on your local drive, beyond just programming.

1. Install Gemini CLI following the instructions at https://github.com/google-gemini/gemini-cli
2. git clone some code repository, perhaps https://github.com/lakshmanok/generative-ai-design-patterns/
3. Issue the command in Gemini to go to that directory

Go to the project generative-ai-design-patterns

4. Ask questions about the repository. Try this:

what is the user interface flow in the app?

Understanding a code repository

4. Make some change to it (I’ve already made this change, so you have to try something else)

Using Gemini CLI to suggest code changes or make changes to your file system

Changes suggested by Gemini

Conclusion

I hope that you tried each of the steps above, and that you tried it until you were happy with the results. What surprised you about how capable the technology is? What surprised you about the errors that it still produces?

Links to Full Series

Day 0: Introduction to series

1. Day 1: Use GenAI as a secretary
2. Day 2: Use GenAI as an executive assistant
3. Day 3: Use GenAI as a librarian
4. Day 4: Use GenAI as an analyst
5. Day 5: Use GenAI as a researcher
6. Day 6: Use GenAI as a sparring partner
7. Day 7: Use GenAI as an engineer (this page)