I'll be explicit: Claude is just another tool in your SW engineer's belt.
If you believe Claude makes you a good engineer and you previously weren't, I am saying that's not true and you still are not a good engineer even with the latest-and-greatest Claude model.
The difference is between "helps" (in your comment) or "you are". Sure, it helps a good engineer do more, do better, etc — but the thread was on being a good engineer.
I was a "good" (whatever that means !) SW engineer long before Claudex. At least good enough that both users and bosses had nothing but praises. And I always took my job and the needs of the users seriously.
It's "just another tool", sure. But one that is so powerful that some things that used to take a day now take minutes, or ones that used to take a week now take a day. And I get even more praises now, along with more time to focus on understanding the needs and controlling quality. For me it's not really about stuffing as much features as possible, but providing better software.
I'm glad this happened after 25 years in my career. I believe I'm in a privileged position where I can benefit from LLMs and still have the knowledge to effectively correct the machine or go back to "manual mode" if anything goes wrong.
Sounds like we are in agreement: LLMs are a great help to an already good engineer.
But also to the original Simon's point that using LLM does not a good engineer make — how we make one is still on us with 10+ years (I've got "only" 20) of experience to figure out now that LLMs are here!
Implicit definition of a good engineer is the one who chooses the right balance between effort, complexity, performance to build a working software system for a business need that can be evolved according to (unforeseen!) future business needs at reasonable cost.
I'll be explicit: Claude is just another tool in your SW engineer's belt.
If you believe Claude makes you a good engineer and you previously weren't, I am saying that's not true and you still are not a good engineer even with the latest-and-greatest Claude model.
The difference is between "helps" (in your comment) or "you are". Sure, it helps a good engineer do more, do better, etc — but the thread was on being a good engineer.
I was a "good" (whatever that means !) SW engineer long before Claudex. At least good enough that both users and bosses had nothing but praises. And I always took my job and the needs of the users seriously.
It's "just another tool", sure. But one that is so powerful that some things that used to take a day now take minutes, or ones that used to take a week now take a day. And I get even more praises now, along with more time to focus on understanding the needs and controlling quality. For me it's not really about stuffing as much features as possible, but providing better software.
I'm glad this happened after 25 years in my career. I believe I'm in a privileged position where I can benefit from LLMs and still have the knowledge to effectively correct the machine or go back to "manual mode" if anything goes wrong.
Sounds like we are in agreement: LLMs are a great help to an already good engineer.
But also to the original Simon's point that using LLM does not a good engineer make — how we make one is still on us with 10+ years (I've got "only" 20) of experience to figure out now that LLMs are here!
Implicit definition of a good engineer is the one who chooses the right balance between effort, complexity, performance to build a working software system for a business need that can be evolved according to (unforeseen!) future business needs at reasonable cost.