He's working 10 hrs. a day, just entering things into Claude, waiting, reviewing, and learning absolutely nothing.
At this point anyone who was truly passionate about programming, is just less and less motivated to become a better engineer.
we shifted to 100% AI gen code a month ago and i realized my knowledge acquisition has nearly frozen. sure, i can evaluate and learn a little when i recognize an inconsistency at the edge or know there are alternatives when an LLM confidently makes a decision without presenting alternatives. but for each instance i catch i assume there is at least one other thing i missed
on the flipside, this sort of means the more ignorant someone is, the faster they will produce. not a problem yet but i think it will be
It's a lot easier to produce "correct" code when they don't have a reference point for what "incorrect" code looks like. The optimistic side of me believes we'll discover new ways to scale websites because of this knowledge gap. A lot of cool tech was built in the 2000s precisely because we didn't know how stuff scaled yet.
I stopped doing it as a result.
If you're telling an agent how to implement something, congrats! You're doing it right and should go on a walk while it's being implemented.
If the agent is telling you how to implement something, you should probably take a walk to digest it, and come back to tell that implementation-machine what to actually do so you can take another walk while it's being implemented.
> A year ago leadership told us everything goes through agents now. Hand written code needed a reason
I can't imagine needing to ask for permission or having to justify myself every time I personally wrote code. That definitely sounds like hell
But in any case, in that hypothetical scenario, just automate your day job and spend time coding a side project you're passionate about. You might have to spend 10 minutes out of every hour prompting, but hey, that's still 400 minutes a day for your side project of choice.
We don't have try to use them as 'agents' to write and understand all the code for us, frankly I wouldn't trust them to do so, they make too many mistakes and foolish decisions about architecture.
Not all of them. Some simply can’t afford it. Some hold true to processes to ensure quality code and knowledge behind contributions.
But what appears to be happening is a race driven by FOMO. Teams are being told to use more and more AI to speed up their productivity. Surely their competition is using AI to build more features, fix long-standing bugs, and make products more efficient. By not spending money on tokens it allows the competition to expand their share of a market. They can’t accept that; shareholders, investors, and owners would replace them with someone who is willing to pull out all the stops.
For startups it makes some kind of sense to take risks in this way, for incumbents it makes no sense.
In my experience, using LLMs to automate boring tasks gives time to learn about the interesting parts.
As to working 10 hours a day, was he working less hours before LLMs, or what's the relevance?
lot of performative but in practice half of the job is looking busy.
Sure pal. All jobs are totally important and not at all pointless, wasteful make-work.
to paraphrase Cat Power: we could all be free... maybe not.
A horse can't do 70mph for 600 miles between feeds. A horse can travel across obstacles that a car would find impossible.
It's horses for courses and a mangled analogy.
If a computer is doing the grunt work, the person should be used to make the best decisions possible. That can't be done while exhausted because you're working ten hours a day.
Queue up longer term work at the end of the medium-length day and go home/log off.
Software development has always evolved.
A preference for it to stay the same way isn't always in our hands.
You can have a say in how software development evolves by how you choose to learn the tools in a way that can influence other people.