When the AI extension came in, I started writing less code, reading more.
When Claude Desktop app came in, I did not like it, stuck with the VSCode extension.
Then the app got better, and the extension got worse. so I switched my focus to that instead of the code editor.
I now view the diffs in the Claude app instead of VSCode, and only switch to VSCode for a more thorough review.
Now VSCode ships with a new view, similar to that of Claude.
I rarely open VSCode now. I rarely write code now.
I miss writing code.
One thing I would note is, I build AI Agents as my job these days, if an agent process is complex, its impossible to write even 90% working code via LLM's.
Is this something people widely vary on? I suspect a lot of people are not like me given that I do see people saying things such as “I rarely use my ide/editor any more” with AI, hence my question to you. Maybe I’m just bad at keeping the code base all in my head.
I do agree with you that I feel that I am coding less myself and I do miss that for sure. But the job demands otherwise.
You only feel powerful when you boostrap a project because you wirte a lots of skeleton code or init core logic.
Nowaday we dont have chance to even write those kind of code I miss writing code too.
1. Prototyper: comes up with brand new ideas; churns out many ideas, most of which don't ship 2. Builder: quickly turns a prototype/idea into production-grade product/infra 3. Sweeper: cleans up the UI, simplifies the code and system, unships, optimizes performance 4. Grower: takes a product that has been built and iterates on it to improve Product-Market Fit 5. Maintainer: owns a mature system to make it secure, reliable, fast, and efficient as it scales
This hits hard. I used to feel a joy and sense of pride in finishing a day.
I am trying to find joys in other ways, like shipping quickly while keeping quality but engineering really has changed in a short space of time.
The understanding derived from actually writing the code and making it work is much different and deeper than looking at code written for you, looking at it, and thinking, "yeah, that seems like it makes sense".
Even if I spend a significant amount of time reviewing the code to try and understand every line that was used, I don't necessarily understand why it was chosen, what were the alternatives, what would the trade offs of those alternatives be, what seems like it should work but actually wouldn't, how something might break if a little part is missing... These are all things that are better understood when actually writing the code.
I care less about the actual code, and more about algorithms, data structures and architectures. I'm glad that AI allow us to think less of syntax and more of logic. My suggestion is that you can use Claude through the terminal and proof read and approve the code, and if time allows it, you can write some of the code you want yourself.
Weird times
There isn’t a single SOTA model on the market that writes code as nicely as I write it.
I can only suspect these people saying they write no code have zero taste, because that’s the only way I can understand accepting the unsophisticated garbage these models produce.
15+ years writing code and refining it, learning. Is AI code as good? Definitely not. Is it 80-90% or the way there with correct linting, agent setup and rules? Absolutely.
It's quality vs speed vs cost adage