[pmarreck]

> LLM watch-six-agents-writing-and-you-proofreading gave me so much existential crisis and depression

this is extremely bizarre because I’m 53, been coding since 12, and it has had literally the exact opposite effect on me, I find it tremendously exciting, like riding a snowmobile instead of manually cross-country skiing

but I do think that if you’re not ready to work like this, you may need to consider a career pivot in the short term

[matsemann]

Your analogy is a bit weird to me. Snowmobile is exciting for a short while, but I'm much more fond of cross country skiing. The connection with nature, the silence.

Or maybe your analogy is correct. AI is a bit as if everyone in the mountains drove around in a snowmobile, noisy and a smell of gasoline.

[cortesoft]

The analogy makes sense. Some people love riding snowmobiles, some people love cross country skiing, and some people love both. It makes sense that some of the people who love snowmobiling think cross country skiing is boring and tedious, and some people who love cross country skiing think snow mobiles are loud and obnoxious.

I don't think people are confused why there are the different types of people who like different winter sports, but people seem shocked that opinions differ on the enjoyment of using an LLM

[qnleigh]

I think the analogy hits home on both sides. You go faster, but you miss the meditative experience of going slow.

[heddycrow]

My knee jerk is that there are quite a few people who can't or won't snowmobile when needed and ski when needed.

That's where the analogy starts to break a bit. You can't mode switch between skis and snowmobile, but you sure can ai assist/not pretty quickly.

One more quick one - imagine skiers showing up to the snowmobile club hating on snowmobiles and vice versa.

I, for one, have still not properly got a grip on how tech enables this sort of a analogy-breaking reality.

Effing go ski then; there's even a club for that! (rhetorical, not directed at anyone in particular) And shame on me cause I show up to the ski club on a snowmobile with skis on my back.

[qnleigh]

Actually I think it still kinda works: - You could ski easy routes but snowmobile harder ones (not that anyone actually does this).

- Snowmobiles are loud and chew up the snow, analogous to AI flooding a repo with low-effort PRs?

[heddycrow]

I can guess that you like this exercise and you seem to be adept at it. I can't help but think you are my kind of people in that regard. Kudos!

[pmarreck]

I think this era has divided developers into 2 types:

1) Those who are interested in the journey

2) Those who are interested in getting to the new destination

[amelius]

Coding since 11. Using AI makes me completely lethargic. I really don't know how to fill the minutes that AI takes to write code. Maybe if AI gets faster I will be able to enjoy it.

That said, like many people here I have invested quite some time in becoming a skilled and experienced coder, so there is no denying that this whole AI craze makes me feel like something is taken away from me.

[microtonal]

I really don't know how to fill the minutes that AI takes to write code. Maybe if AI gets faster I will be able to enjoy it.

I either switch between two projects, or I keep an eye on what Claude is doing, because it often gets off the rails or in a direction I don't like and then it's easier to just stop it there and tell it what to do instead.

That said, like many people here I have invested quite some time in becoming a skilled and experienced coder, so there is no denying that this whole AI craze makes me feel like something is taken away from me.

I might have felt like that when I was younger (almost 44 now, programming since 10), but over time I realized that the thing I enjoy is not really writing code itself, but coming up with ideas, solving puzzles, etc. LLMs are like insanely fast junior programmers, so they do the more mundane part of the task, but they need me to come up with good ideas, good taste, and solve design challenges. Otherwise it ends up as a pile of unmaintainable junior programmer code.

It is possible that LLMs might replace the other parts of being a good programmer as well, but for the time being it makes my work more pleasant, because I can work on interesting problems.

[zozbot234]

> I really don't know how to fill the minutes that AI takes to write code.

The AI should be spending most of its time helping you spec out new revisions to the codebase, the code-writing time is just the last step and if you've planned the work in depth, you'll understand what the AI is trying to do (and be able to stop and revise if anything is going off the rails). This is a healthier approach than "just spec out something else in the meantime" IMHO, but of course that happens too.

[ineedasername]

"just spec out something else in the meantime"

Yeah, I've learned that if I do too much of that I'll spend more time catching up in terms of consolidating gains through review of code and functionality. That's just me, people are clearly developing a few different and not "wrong" ways of going about things.

[tarr11]

52 here, been a full time people manager for about a decade now. Coding manually makes me tired just thinking about it. When I think about embark on a new project my mind goes back to all the times I worked 12 hour days trying to get some basic system to function. I’m too old for that now, my back hurts if I sit too long and occasionally get migraines if I look at a screen too much.

Using AI has been really perfect for me. I can build stuff while I do other things, walk the dog, make lunch, sit on the porch.

Sometimes i realize that my design was flawed and I just delete it all and start again, with no loss aversion.

[sidrag22]

> Using AI has been really perfect for me. I can build stuff while I do other things, walk the dog, make lunch, sit on the porch.

this resonates with me strongly, while i like coding, and understanding it, I understand my human limitations. I couldn't possibly write by hand the stuff I've been making, in the time I am making it, without a team these past few months. I would be coding literally all day, which while I sometimes enjoy the zoning out process of wiring stuff up, what i really enjoy is exactly what you described.

I enjoy being outside and walking my dog, taking a long shower, and cooking. All of these things are simple tasks with a good bit of repetition, and unlike wiring up some code or whatever, they allow my thoughts to flow, and I can think about where my projects are likely heading and what needs to be done next.

Those moments, even before heavy AI assisted coding, have always been the moments i cherish about software development.

[gdilla]

For me, coding since the 80s (but i knew then it didn't spark joy or anything - debugging was so annoying, learning new language syntax even more so...) I love AI. I am a product manager, i just see freedom to make things that are real and learn faster - does this solve a problem? Is it better than what we have now? and move on, disposing of things as i go because it's cheap. To fill the minutes, i might work on 3-4 or even 5 separate projects and even multiple worktrees within those. I feel busier than ever. I think the best part is it's not lazy and I am. There's so many things I don't have time or energy to go deep on that I can delegate. I'm jealous of real sw engineers because it's probably a huge force multiplier, while I can't call BS on it as much, but getting better.

[igravious]

two AIs -- I use Claude Code and Kimi CLI -- I got them to build an agent relay so they can communicate with each other (one plans, the other reviews the plan; one builds, the other reviews the build) -- while one is working on one thing, I'll be chatting and exploring with the other one … they can build anything in any language so if you are a skilled and experienced code you should be able to guide a pair of coding agents no problem.

Otoh -- if there is this bifurcation among coders (one group super-excited, one group depressed and angst-ridden) then maybe we should be trying to figure out why people are reacting the way they are. Can you explain more about your situation? What do you code? Do you have hobby projects? Do you have free time? Etc.

[ineedasername]

"how to fill the minutes that AI takes to write code"

I usually review the code that's been written. Sometimes directly, sometimes by telling claude to bring things up piece by piece to explain choices as I review. Or I kick off one of the various maintenance tasks, validate my assumptions and expectations on how things should function, note the things that don't to be addressed. I'm going to have to do this stuff anyway, I might as well do it then.

Or I read something, or do something to clear my head. Sometimes because I need a mental break, because I find that the speed these tools having me working at can be taxing in different ways.

I think expectations of the "10x" variety whether you put that at 10 or 3x will have to be adjusted. Coding as fast as 5 developers is far far different than "A singly developer can produce as much as 5 others"

[heddycrow]

+1 review the code

Screw {some number}x. Such BS. Those who can, do. Those who can't, write and spread pseudo-intelligent brain worms. Reject!

[ngcazz]

I live with a feeling of nonaccomplishment, never having taken a project to completion thanks to my shitty executive function. The AI craze has robbed me of any hope that I might still meaningfully* achieve this in my career.

[axus]

Couldn't an AI help you take a project to completion, at least until you run out of money?

[ngcazz]

I didn't mention the crucial point which is that I what I signed up for was writing my own software.

I grew up witnessing Carmack going from Keen to Quake in 5 or 6 years.

That standard gets you attached to the idea that you should be able at some level to individually reach a fraction of the depth and breadth. Sadly, I have neither the energy nor the focus.

But what's the point of getting an LLM to, say, write a raycaster if your point is to learn how to do that yourself? If your mission in life is to learn to build things?

(I hope I'm getting my idea across)

[ygjb]

You have conflated the joy of learning with the joy of building. I have been writing code since I was 6 years old and was left to my own devices with the vic-20, the manual, and BASIC instructions.

I have worked as a developer, security engineer, program manager and engineering manager through my career. Writing stuff to understand algorithms or hardware requires engaging with the math, science, and engineering of the software and hardware. Optimizing it or developing a novel algorithm requires deep comprehension.

Writing a service that shuffles a few things around between stuff on my home network so that I can build an automation to turn down the lights when I start playing a movie? Yeah, I could spend a day or two writing and testing it. Having done it a few times, the work of it is a bit of a chore, I'm not learning, just doing something. Using an Claude or some other agent to do that makes it go from 'do I want to spend my time off doing a chore?' to 'I can design this and have it built in an hour'.

Making the jump to using the tools in my day job has been a bitore challenging because as a security engineer I have seen some hairy stuff over the last two years as AI generated stuff wends it's way into production, but the tools and capabilities have expanded massively, and heck, my peers from Mozilla just published some awesome successes working with Anthropic to find new vulns :)

Don't let using tools take away the love of learning, use them to solve a problem and take care of the drudgery of building stuff.

[heddycrow]

OMG that manual. VIC-20 was my first code experience. I look back and cannot understand how 7 year-old me was patient enough to make a jumping jack guy appear on screen. Joy of Coding? Hell, no. I wanted to see if I could make it work. (I did, and I had no clue how to save to tape)

Sounds like you had one at home? If so, I'm a bit jealous. But also, hello, brother/sister!

[ngcazz]

I appreciate your thoughtful response, you might be right, maybe there's an element of getting over myself to stay in the race...

[Aeolun]

The trick is to have 3 terminals with Claude Code open at the same time. You won’t be able to follow more. Reviewing the stuff or plans written is harder than telling them to write it.

[pmarreck]

> I really don't know how to fill the minutes that AI takes to write code

At any given time I have at least 4 terminals working on various pieces of a main project and/or side projects.

Whenever I have to wait on 1, I just flip to another. I find it very engaging because I can work on multiple pieces that all contribute to some overarching goal at the same time, concurrently.

[jcelerier]

> I really don't know how to fill the minutes that AI takes to write code.

Think of it like being a project manager in a team. There's a lot you can do to help the project moving forward without touching one line of code

[silh]

I would’ve become a project manager if I enjoyed that, right?

[jcelerier]

no one's preventing you to write the code yourself

[gmadsen]

Multiple projects at once or at least multiple features. I am usually the limiting factor reviewing, not waiting on agents

[ay]

I am 50, coding since ~12. Started with Apple II, during the uni times wrote my own editor in assembly for BK-0010 (a soviet computer), then 30 years in computer networking with some high performance dataplane stuff more recently;

The last years somehow it felt like there’s nothing new anymore, the same 10 ideas being regurgitated with slight modifications. I tinkered with AI for the past 2 years but it was mostly a “tool for writing boilerplate”. I have tried a few ideas for agents but didn’t see how it could work.

That changed with Opus 4.6 and the subsequent wave of local models - now I try 10 ideas a day and it’s like magic! And if something doesn’t work - jumping into the code and debugging it is huge fun!

Understanding that the era of the almost-free cloud tokens might come to an end, I run my own harness pointing to my own GPUs running Qwen3.5-27B, and the last few days it has been very busy! :)

My harness doesn’t “pressure cook” since it doesn’t make sense to do that with only one GPU (besides many other reasons), it runs everything in a linear fashion, including subagents, and logs everything - reading the logs as they go by is another cool thing - sometimes I pick up interesting things from it !

The distribution of people’s moods related to AI seems indeed bimodal. And I feel lucky somehow ending up in the “enthusiastic” rather than “depressed” part of it. To the folks in the other one: I am sorry. I don’t know why it is this way. If I knew I might have given unsolicited advice.

[techpression]

So you’ve tried at least a hundred ideas by now, care to share fifty of them? I’m very curious as to what they are. Opus is too slow to even complete one idea per day for me, and that’s fine, I don’t have hundreds of them :)

[ay]

I dont have big ideas. Some of the more interesting ones that I ended up using but can’t share: a streaming radio for my MP3 collection (runs behind the vpn); a lightweight and self contained webrtc conference server for talking with my family; a process-level virtualization based on KVM.

Of the ones I can share:

Browser-based network tester using webrtc unreliable data https://netpoke.com - use magic code “DEMO” to see what’s it about - the source is at https://github.com/ayourtch/netpoke

A port of the SOTA speech generation model from Python to Rust:

https://github.com/ayourtch/fish-audio-experiment

A study on LLM prompting techniques:

https://github.com/ayourtch-llm/kindness

My own coding agent that i use with my locally hosted LLM for experiments:

https://github.com/ayourtch-llm/apchat

Also LLM helped with a lot of code for my packet mangling library: https://github.com/ayourtch/oside - which, among other things, includes a now battle tested SNMPv3 stack.

A true “stochastic parrot” using hash tables: https://github.com/ayourtch/hashmem

These are the ones I remember. Feel free to scout my GitHub for more. Edit: And of course it doesn’t need to be said that out of ideas I try all of them make it to github. Many end up thrown away.

[ChrisMarshallNY]

I’m retired, so a career pivot isn’t in the cards for me.

I’m also not really in the HN gestalt, so to speak. I have some views that are common, hereabouts, and some, not so much.

I’m enjoying having an LLM “pair partner.”

[HoldOnAMinute]

Yes, it's a partner you can discuss things with, and it never belittles you or gets tired.

[krd8ssb]

I'm with you on this being incredibly exciting.

I'm 40 and have been doing this since I was 12 as well. Once I became a staff engineer at a large company and ended up being a less Hands-On with code and more on team leadership and system architecture, it set me up for this perfectly.

I missed writing code (or so I thought) but what I realized is that I actually missed bringing ideas to life. Coding was just a means to do that and the new tools with LLM and agents have allowed me to do the core of what I love way more than coding by hand could have ever allowed

[accounting2026]

Same boat (though 44M) - I don't think it has become less fun, on the contrary it can help with the stuff that was trivial but could still take time to get right. Now it can crank out that stuff often correctly on first try. Of course I have the same fear of job security as everyone else and it is sad to see something you were good at being taken over by machines, but it is not because I enjoy the work itself less, quite the contrary.

[witx]

It's so bizarre I see everyone happy using llms trained unethically, so they stop doing the only creative and interesting part of the whole SDLC and just become managers

[pmarreck]

There is no strong argument that they were "trained unethically". You can google or ask an LLM to argue pro/con that claim... but of course, you haven't, since doing so would itself violate your apparent worldview.

> so they stop doing the only creative and interesting part of the whole SDLC and just become managers

Bullshit (and, speaking as an old, incredibly whiny). Figuring out the right way to talk to yet another API is not the "creative and interesting part." Coming up with entirely new concepts and building them faster than I ever thought possible given that I have zero time to myself since I have a kid is literally the most "creative and interesting" thing I've experienced in my entire life. /shrug

I don't know what to say to folks like you anymore other than "byeeeeee, got too many ideas I can build now to waste time arguing with people who are simply not getting it... and maybe never got it in the first place because they got pushed into STEM instead of being cursed from the gods with it like I definitely did"

[witx]

There is plenty of proof on how these models were trained, you're just too enamored with your "more code, move faster" button to care.

That's fine no one's perfect we all look away from some things here and there, me included. But don't throw shade

[skeeter2020]

Way to blame the poster for their problems, and what the hell is a "short term career pivot"?

[pmarreck]

Who is blaming anyone?

Does someone blame the manual loom operators when automated weaving comes out? The elevator operators? The carriage drivers? The milkmen? No. So please stop reading your preconceived grievances into what I said.

I just see this as something that is not going to go away so the only rational choice is to get ahead of it and master it before it runs you over. If you love the act of creation and can adapt to a certain amount of change (or even see it as exciting), you'll be fine at this. If you have emotional hangups, well, that's irrational (no blame or judgment implied, it just... is literally irrational), and you might have to make some difficult choices that sit better with you.

I'm well-familiar with making irrational choices. I left a well-paid Microsoft shop and joined a startup and took a pay cut just so I could work with open source, years ago. It ended up paying off.

Do what your gut says. It's making the bets, after all.

And guess what. The people who have stared at code for years now are the ones who best stand to gain.

[infecto]

I don’t think anyone is blaming but it’s hard to ignore the progress we are seeing and thinks that these workflows will not be the norm in instead of the exception.