If you are a programmer using an AI LLM to help you code, are you finding it speeding you up or slowing you down? What impact has it had on your programming?
Last Updated: 21.06.2025 00:40

I recently used AI to make this Chrome extension. The chrome extension works like this.
You also get a notification on how much it costs.
Yeah I heavily depended on AI to make this Chrome extension (with Cursor). I had no experience in building a Chrome extension before this. The code is here
When should I use the best sunscreen for oily skin?
I want the extension to be activated when the extension icon is clicked
I want the extension to send a desktop notification when it is complete
Ultimately, you need to decide how much do you need to learn. It is okay to delegate to AI if you don’t intend to learn the things you don’t need to learn.
How can Substack help creators with their overall business strategy?
First you write in human language what you want to do
It calls the OpenAI o1-mini to prefil the cell
I want the extension to be activated on Cmd+K
When I solved one issue (usually by moving code from one place to another), it usually breaks another issue. I have been stuck in these cycles many times. Moreover, testing Chrome extension is a manual and tedious process that wasn’t automated. I think I spent more time using AI to make this extension than if I had learnt how Chrome extensions are being designed and make informed edits.
I want to minimize the requested permissions where possible
I binded Cmd+K to initialize generation
I want the extension to read the storage for the API key
Even though the interface looks simple, it was not easy to make all of the features work at once.
I want the extension to read and write content of the current page, specifically the `Jupyter.notebook` object
What are the biggest challenges in training multimodal AI models?
I want the extension to bring the user to the option page if the API key is missing
However, if there is AI that is able to make this Chrome extension for me end-to-end with testing and minimal feedback provided, using AI here would have saved me time. This is not yet the case.
I no longer need to browse the Internet to find out the certain Trino keywords, or how to aggregate a pandas DataFrame. It also works on wherever you have Jupyter Notebook. I wrote the Chrome extension, the data is only exposed to OpenAI and probably Google.