You’re totally right, but I feel LLMs are a double-edged sword. In my own experience, they work as a creativity booster, but I also feel the problems. That’s why I started my news channel: Singularity News. I also made a card game with an LLM, based on my hand-coded game, which tries to communicate the problems we’re facing.
A Software Engineer who strongly believes you can add immense value by knowingly implementing Software Engineering best practices even at the age of AI, and puts human life and dignity above all else.
To keep your agency (freedom, independence, authority, rights).
This is so important and people always overlook it. AI is completely under the control of a few large companies, often even under a lot of influence by individual megalomaniacs.
I think Frank Herbert really said it best way before AI was anything other than a fantastic sci-fi concept:
“Once men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them.”
― Frank Herbert, Dune
A Software Engineer who strongly believes you can add immense value by knowingly implementing Software Engineering best practices even at the age of AI, and puts human life and dignity above all else.
The agency point makes me think the useful exercise is choosing the problem and its constraints, not just writing the implementation. Even if a model produces perfect code, deciding what should exist, what should stay out, and which tradeoff is acceptable still needs practice. A small project could keep those decisions human while delegating some typing. Would you count that as the deep work you want to preserve, or is implementing it yourself an essential part too?
A Software Engineer who strongly believes you can add immense value by knowingly implementing Software Engineering best practices even at the age of AI, and puts human life and dignity above all else.
Implementing it yourself, or being able to do so and doing that alongside LLMs is essential.
It is not the end product and some intermediate decisions that generate your own neural connections, it's the process and struggle that does it.
It's similar to Gym workout. We humans used to do heavy labor intensive work. Our genetics demand that. Nobody could effort to be obese. Now, that hard work is replaced by workout, running etc. But you yourself will have to do the heavy workout to be in shape. Some decisions, a little workout will not help.
Intellectual ability is no different. If you don't continue to do intellectual heavy lifting, you'll become intellectually obese (i.e. dumb).
The same applies for agency. If we have to depend on AI big companies to do our work, we lose our freedom bit by bit. However, if we use these as our "workers", not as our replacements, then we can still do everything those AI bots/agents do (we use them for acceleration/productivity only), so AI big companies can no longer enslave us.
Fair pushback: choosing constraints alone leaves out the practice of turning them into working code and finding where your understanding breaks. I would make that implementation work an explicit part of the exercise, then try a variation without the assistant to see what transferred. The workout analogy makes sense there, though I would separate the need for practice from the stronger claim that using assistance necessarily reduces ability. What would you use as a practical test of retained independence?
A Software Engineer who strongly believes you can add immense value by knowingly implementing Software Engineering best practices even at the age of AI, and puts human life and dignity above all else.
I think an agentic workflow where you assign features to AI agents, just like you would to another developer as a project lead, along with controlling the core architecture yourself, and implementing some features, and bug fixes yourself without AI assistance at all, is a consistent workflow that'll retain your skills, independence and will require no additional tests.
You have a fair point, though I think people are mostly just tired of AI-generated posts and videos and crave genuine human connection. When it comes to software, however, nobody really cares whether a tool was written by a human or an AI—as long as it solves a real problem sustainably.
A Software Engineer who strongly believes you can add immense value by knowingly implementing Software Engineering best practices even at the age of AI, and puts human life and dignity above all else.
You have a fair point too, however, it's not true that "nobody really cares whether a tool was written by a human or an AI". I do care, and I think, for serious software like banking, health care etc. people should care.
AI testing storyteller. Writing the 36 Stratagems series — ancient Chinese military tactics meet modern AI system meltdowns. 16yr QA → building AI test framework.
A Software Engineer who strongly believes you can add immense value by knowingly implementing Software Engineering best practices even at the age of AI, and puts human life and dignity above all else.
You’re totally right, but I feel LLMs are a double-edged sword. In my own experience, they work as a creativity booster, but I also feel the problems. That’s why I started my news channel: Singularity News. I also made a card game with an LLM, based on my hand-coded game, which tries to communicate the problems we’re facing.
That's absolutely true. It's already extremely useful. That's the very reason we'll have to be careful with it.
This is so important and people always overlook it. AI is completely under the control of a few large companies, often even under a lot of influence by individual megalomaniacs.
I think Frank Herbert really said it best way before AI was anything other than a fantastic sci-fi concept:
So true. This will be a kind of enslavement like no other enslavement in history. We must resist.
The agency point makes me think the useful exercise is choosing the problem and its constraints, not just writing the implementation. Even if a model produces perfect code, deciding what should exist, what should stay out, and which tradeoff is acceptable still needs practice. A small project could keep those decisions human while delegating some typing. Would you count that as the deep work you want to preserve, or is implementing it yourself an essential part too?
Implementing it yourself, or being able to do so and doing that alongside LLMs is essential.
It is not the end product and some intermediate decisions that generate your own neural connections, it's the process and struggle that does it.
It's similar to Gym workout. We humans used to do heavy labor intensive work. Our genetics demand that. Nobody could effort to be obese. Now, that hard work is replaced by workout, running etc. But you yourself will have to do the heavy workout to be in shape. Some decisions, a little workout will not help.
Intellectual ability is no different. If you don't continue to do intellectual heavy lifting, you'll become intellectually obese (i.e. dumb).
The same applies for agency. If we have to depend on AI big companies to do our work, we lose our freedom bit by bit. However, if we use these as our "workers", not as our replacements, then we can still do everything those AI bots/agents do (we use them for acceleration/productivity only), so AI big companies can no longer enslave us.
The following video explains it well:
Fair pushback: choosing constraints alone leaves out the practice of turning them into working code and finding where your understanding breaks. I would make that implementation work an explicit part of the exercise, then try a variation without the assistant to see what transferred. The workout analogy makes sense there, though I would separate the need for practice from the stronger claim that using assistance necessarily reduces ability. What would you use as a practical test of retained independence?
Good observation.
I think an agentic workflow where you assign features to AI agents, just like you would to another developer as a project lead, along with controlling the core architecture yourself, and implementing some features, and bug fixes yourself without AI assistance at all, is a consistent workflow that'll retain your skills, independence and will require no additional tests.
You have a fair point, though I think people are mostly just tired of AI-generated posts and videos and crave genuine human connection. When it comes to software, however, nobody really cares whether a tool was written by a human or an AI—as long as it solves a real problem sustainably.
You have a fair point too, however, it's not true that "nobody really cares whether a tool was written by a human or an AI". I do care, and I think, for serious software like banking, health care etc. people should care.
🤣