One of the hardest things with AI coding (for me, anyways) is finding that balance between pushing forward and truly learning new things.
Not talking about code reviews etc here. I’m pretty happy with my current approach towards that.
But when working with some library, API etc that I haven’t worked a lot with, it’s hard to truly learn it.
Understand it, yes, sure. But since I didn’t have to write all the code from the ground up, struggle and fail, it’s not really „learning“ that’s happening there.
The obvious solution may be to just not use AI until I have learned something (though when are you really done learning)?
But that gets into the way of quick prototyping - and one big plus of AI is that you can quickly experiment and discard.
You could also determine that you don’t need to really understand or learn the underlying technologies, frameworks etc you’re using.
That sounds like a bad idea once you leave prototyping.
Using AI to discuss the code works pretty well. It’s what I’m currently doing - in addition to traditional forms of learning. Docs, tutorials, YouTube guides - that all is really helpful when combined with AI for discussing the code and implementation.
Has the positive side effect that this reflection about the code can also help improve it.
Anyways, still figuring this out. What works for you?
Introducing Claude Sonnet 5, our most agentic Sonnet yet.
It makes plans, uses tools like browsers and terminals, and runs autonomously at a level that just a few months ago required larger and more expensive models.
Today we're releasing a new set of components for building chat interfaces.
We've taken the patterns we build every day, rethought the abstractions behind them, and turned them into components you can compose and customize.
We're starting with the conversation layer: streaming, scrolling, messages, bubbles, attachments, and markers.
The essential engineering cheatsheet of 2026:
agent → while loop
subagent → nested while loop
agent harness → the rest of the code
cloud agent → all the above, on EC2
⚡️ Vite 8.0 is here!
The most significant architectural change since Vite 2.
⏬ Powered by Rolldown bringing faster production builds and more consistency
🛤️ New features such as tsconfig paths and emitDecoratorMetadata support
i have lost meaning. nothing feels difficult anymore.
earlier, there was so much joy in writing code by hand - finding problems, staying awake for hours just to solve that one silly bug buried deep in the code.
coding gave me a lot of joy, but with agentic coding, writing code by hand no longer makes sense - not because i want it that way, but because of the pressure from everyone around you. why struggle for hours when you can finish everything at 100 toks/sec?
it has left a dent in that joy i once carried. i wish i could go back to writing pure code with my own hands - no agentic coding at all - but there is no time for that. features need to be shipped quickly, and if you don't keep up, you get crushed by others who do.
some might say, "you still have to read the code and figure out if it's correct." that's true, but reviewing something is very different from actually writing it. the neurons fire differently for both - and they fire far more when your fingers are hitting those keys on the keyboard.
maybe the craft isn't dead, but it is slowly being asked to wait in the corner - and for those of us who fell in love with the process, not just the output, that silence is louder than any compiler error ever was.