Hacker News Reader

Comment by weatherlite | Hacker News Reader

My nightmare scenario (which might start to materilize) is that our last years in the industry will be becoming prompt monkies / agent "managers" working on codebases we barely understand in such velocity there's no way we can gain real understanding. Whenever something breaks (and it will , a lot) A.I will fix it - or so we'll hope. And the sad thing is - this might work; you'll get more stuff done with fewer people. Sure, we didn't sign up for this, it's not a fun job what I've described, but why should management care? They have their own problems and A.I is threatening their jobs as well.

View on HN

At work we build enterprise software with stuff like Kotlin+Spring + multiple NextJS apps + Microservices + Rust CAD engine.

I haven’t have written code aside from tweaking stuff here and there in probably 3 or 4 months. Before that I wrote code by hand every day for many years.

I’ve found a lot of fun parts of my new workflow that I enjoy. I still miss being fully immersed in a problem deep in the files… and sometimes it feels like homework reading so many implementation summaries from Claude because the feature spans 4 repos and is too much code to read. But I do love shaping the code into different solutions exploring in a way that is unique to ai native workflows. And I love building agent skills and frameworks with/around them and expanding it out to more aspects of the company or life — there’s deep work to be had that still feels like hacking in the trenches. I get a lot of the same satisfaction in different ways, and there’s a lot of exciting novelty to explore that was previously out of reach due to time and energy constraints.

Also I don’t like our backend stack and I hate React / NextJS to the degree of derangement syndrome — I am so happy that I don’t have to write it and I can just focus on UX, making customers happy / lives easier / shaping the software into better and better versions of itself at such a faster pace.

People who learned good software engineering intimately before the inflection point are extremely lucky right now. Existential dread and the stages of grief have been a part of the journey for me too sadly, but there’s a lot to celebrate and explore with the right attitude.

I feel the same way, I have many years of experience, and I have gone from writing everything by hand to using claude code all the time (my latest company is very pro doing everything with AI).

Since I have been a software architect for the past 7-8 years it feels in some ways that that experience makes using claude code a lot more productive than for my non-architect colleagues, as I am able to steer it much more effectively whether directly in sessions or via custom skills / mcp.

The big issues right now for me are hiring and manager expectations, I changed positions last fall due to mass layoffs and it took me 3 months to find one: having leetcode interviews in the current climate seems completely useless, even more than it was in the past, and system design interviews are so formulaic it also feels like a crapshoot. Plus every job getting hundreds of AI generated applications makes actually being considered in the first place quite difficult.

Manager expectations are also ridiculously inflated nowadays, it seems most action items that come are claude written with fantastical random statistics (if you add caching you can make your backend 98.3% faster!), and it takes so much time to fight this and unrealistic team velocity expectations.

Interesting times, I do feel lucky I have had a long career, but I very much fear the ladder being pulled up even more than it has been when outsourcing because widespread. I know everybody says "things always change, new opportunities will open up to compensate for the ones that are being lost" but this time it does feel different, and not in a good way.

Things are changing so fast and so chaotically with this technology. I'm also writing everything now using Claude code, and I've been thinking a lot about what this means for my work moving forward. One thing I've noticed, is that I will just keep hammering and hammering on my work until I force myself to quit. Even on the weekend I feel the pull to go work on it. I'm just less sort of mentally exhausted by work, I suppose, but I don't think that's particularly healthy if it leads to me working way more than I should. On one hand, I think that's a reflection of how powerful and exciting this technology is, but on the other hand, I think that it triggers some different kind of reward function in my mind that I'm not used to.

In any case, I think if one wants to continue to have a career in this industry for years to come, it's basically table stakes to become fluent in using these tools.

Similarly, I started using Claude to add some features to an native app on iOS and Android in early January. That was so successful (tooks days instead of weeks) I started applying it to client work and basically haven't written any substantial code since. A big change from around 40 years of writing code pretty much every day and I'm enjoying the increased velocity from not having to web search for API and CSS syntax details.

My son, working in another dev company, reports the same - he basically hasn't written much actual code for about three months. It's a massive change.

It already happened. The old timers correctly observe that modern applications are bloated and inefficient because of all the heavyweight frameworks, excessive abstraction layers, and "left-pad culture" where external dependencies are pulled in to do the most trivial things but that these things enabled less capable developers to effectively build software to fulfill industry demand. LLM-only coders are just the next step in the devolution.

> My nightmare scenario (which might start to materilize) is that our last years in the industry will be becoming prompt monkies / agent "managers" working on codebases we barely understand in such velocity there's no way we can gain real understanding.

It will always be preferable to work on an understandable codebase, because that maximizes the AI's affordances too. And then the AI can explain things to you. A skilled human will always have a lot of solid knowledge relating to their hyper-specific niche that isn't part of your average general purpose AI, so humans will obviously have a key role to play still.

That's literally what https://news.ycombinator.com/item?id=47157039 seems to be about, and I had the same reaction as you

I'm already seeing this in the company I recently joined: 80-90% of code is generated/prompted. Big PRs, very little review or oversight. Absolutely nobody considering long-term architecture (and IMO nobody capable of such). In general, there's very little critical thinking involved at any stage, just throw error messages back into the LLM, rinse and repeat. I'm hoping there's a world where people with skills are useful in getting these projects back on track, but perhaps as a society we're learning to accept this reduction in quality.

And how do u sum up the tradeoffs so far, or is it too early to tell ? Do u see lots of unacceptable shit making into production that wouldn't have before A.I for example ?

Observations so far:

- "micro" level code is adequate. A new feature typically works as expected, and code is well documented. But macro level everything's wacky: no architectural consideration, lots of duplication, dead end paths, dead code. Nobody ever took the time to understand the context surrounding a change and figure out the most appropriate way to integrate it.

- there's no consideration for infrastructure, or how the code is run (my world). Last week we ddos'd ourselves after someone implemented a 1/s poll that runs on every open browser tab, worsened when the frontend's reconnection logic fired 1/s with no exponential backoff.

- zero critical thinking... First thing everyone did during that incident was to ask their LLM of choice why the platform was down. Investigating by hand (and my findings thereof) was dismissed, because surely Claude knows best.

I'm too old for this stuff.

Your codebase is essentially becoming bytecode

Reply on news.ycombinator.com