I'm getting more done than ever, but pride in my work is at an all-time low. The average quality of the work I do has improved dramatically over the last 6 months. Many of the engineering disciplines I neglected or never learned happen automatically now. It used to be I would sweat to make something I was proud of, because I believed it was worthwhile to do so. The bad parts were bad, and many, but eclipsed by the purposefulness of the work that drove me to do it. And the good parts were good, because they were real — because I cared enough to discover them. I'm not espousing the labor theory of value. I'm not talking about some essential, ineffable "good" in craftsmanship. What I'm saying is that sweat is an essential ingredient in producing something with a soul: something that people can love. People don't love commodities, they love singular things. The commodification of software is killing the love of it. In other words: other people loved the garbage I used to craft more than the garbage I now produce. Nowadays, all the parts of my work that used to be low quality are high quality, because Claude is a better programmer than I am. But the parts of my work that used to be good haven't survived. I might be trying to do something worthwhile, but because I haven't actually engaged with the problem I can't make anything better than slop materialize. The glory has departed. Flotilla has greatly profited from me using AI to code, because the good parts only I could build were already there, and there was a lot of bad stuff that needed to be fixed. But the way I feel about other projects that I haven't gone through that journey with is that they are garbage. Total trash. They are as valuable as the tokens I used to generate them. Because there's nothing really there. There is nothing to love. There is nothing singular. They are disposable, fungible. When people refer to themselves as "idea guys" it means they suck at doing real work. But their ideas can only exist if someone takes the time to make their ideas a reality. But in the New Age of AI, idea guys can finally realize their genius, without being held back by a bunch of narrow-minded technicians! But I think it will turn out that there wasn't actually any genius there in the first place, and everything that was of value about the original idea was actually discovered during the process of building it by the people who did the work to make it a reality. Software as an art, or a craft — as a thing which is an object affection — is dying. This is fine, the affection was probably misplaced in the first place. A new kind of software — the stuff we create for ourselves, that no one else will ever see — will probably absorb some of that affection. The rest will be commodified. I think this goes for every industry, by the way. An accountant can feel proud of his spreadsheet, a plumber of his solder, a salesperson of his pitch, even a developer of his agent harness. He can love the work of his hands and have affection for it. It can be singularly valuable to someone, and a conduit of love between the people concerned. Does an assembly line worker feel affection for the parts he stamps out, or the people who receive them? Is oil extracted with heart? Are the foundations of skyscrapers filled in with love? Does the dollar store worker bask in the sunshiny faces of the children purchasing their very own extruded plastic disposable pseudo toys? Maybe they do. I do think that wherever humans use tools, there will be affection. It reminds me of the nose art pilots used to paint on their bombers. "This is my rifle. There are many like it, but this one is mine." Maybe means I should buy a mechanical keyboard, I don't know. DAU and MRR probably engender warm fuzzies in some types of people, but I'm not one of them. Commodities are not necessary for a good life. A community can live without commodities, and survive. If everything were hand-made, pride and affection would ubiquitous - and its absence would be traceable to a character flaw in the worker. If every meal came from an animal who had a name instead of a number, every meal would embody gratitude and grief like no store-bought meal could. I yearn to live that life and check out of this Terry Gilliam-esque Hotel California life, I say as I glance out the side of my eyes at the people I know who have just bought a couple dairy cows and moved out to some acreage, and keep typing. Am I a coward for wanting to stay where it's comfortable and secure? Where my skills (or what's left of them) are in demand? Is my heart too bound up with the bounty of the market that I have become a commodity myself? Or am I a whiner for being so sad about losing the drudgery of programming? For not wanting to live on a diet of petrochemical-derived macro-nutrients? For not wanting to spend hours a week in a car driving to an office where I work by myself? "Better a handful with quietness Than both hands full, together with toil and grasping for the wind."

Replies (47)

Want to listen instead of read the screen? I'll turn it into audio for the thread once the 918 sats bounty is reached. Total can be from one zap or many.
The pride from delivering something of value (regardless of 0 cost of actually grudging it into existence) is still there. Think of all the things you would simply not be able to deliver without losing a (real) life, but now is simply possible. I'm having a blast.
Anyway, thank you for sharing this @hodlbod. I am having a similar emotional journey, albeit much shorter. 3 years ago I didn't know much about programming, and my first 2 years were quite rough because I didn't delegate anything to the clankers, but actually used AI to teach myself and was questioning everything. In hindsight that was good, and the pain I felt when learning and failing and finally making it right played a huge role in making me going to sleep tired but with a sense of satisfaction. However I felt a similar kind of satisfaction after doing some manual work like placing a parquet, painting the walls and similar. The sweat creates the meaning. For software engineers to find a similar sense of satisfaction I see two roads. The first one is dreaming bigger, being bolder, attacking previously unexplored problems, maybe even ones were the success is tied to human understanding itself. As an example, I am researching if I can take Leela (an AI that play chess, similar to Alpha Zero if you know it), feed it some positions, look at the internal representations and extract chess concepts that may be novel, and use it to create an AI chess coach or something. If this shit works, I'm pretty sure I'll find satisfaction because it's an incredible project that no one has ever achieved (compared to may standards). Deepmind did good work on it, and is still doing, but not even close to my goal. And without AI there would be no fucking way I can attack this problem, I would just lose years on small implementation details where instead I need to iterate quickly. The second road is completely different, and it's custom made software. Now it's actually possible for a small business to outsource a piece of software they need, made to fit their need perfectly. In the past this would have been way too expensive. Today in fact I showed a friend of mine who has a medium size business the website + we app demo I made for him, and man, sawing his face in awe gave me a lot of satisfaction. The opposite of the commodification of software, this is software becoming craftsmanship, where there is one customer, not many! I'll make this for you, and because of that I can perfectly map it around the way you work, they way you think, your values, your brand, everything! A custom digital suit. A software artisan. I'm trying this road for some time, I'll report back.
open jets > opensats gigi open wallet and send jet dont forget the bacon
When I read what you wrote, I immediately thought about the consequences of the division of labour due to the Industrial Revolution, and how this affected people’s wellbeing. But in that case was a mechanisation of physical production and everything around it. This is different because is mechanising cognitive production and that is why it feels so invasive. You don’t need to understand to produce and it could be scary. My strategy has been to set limits on what I do with AI and to choose aspects of my work that require less understanding and judgment on my part. As for the rest, I handle it myself. I might occasionally have the AI ​​review it for logical consistency, just in case I’ve missed something. Ultimately, however, the final decision on what to accept or reject is always mine. Does this make me less productive? Perhaps, but I retain essential skills that extend beyond the scope of my job.
AI can make the work more efficient, but efficiency isn't the same as fulfillment. Sometimes the struggle is where we discover what makes the work meaningful. Maybe the real challenge now is learning how to use these tools without losing ourselves in the process. View quoted note →
I usually dont read long posts like this, but I read yours. I like your honesty here. I have nothing to add but would love to hear you and Gigi talk about this for 3 hours 😉
Resonates here. Those that are “excited about it” are usually non-programmers that now can build something by themselves. This will surely fade as the novelty wears off. Those that say you can now build something above what your skills allowed before, fail to notice that if you build something you don’t understand, you cannot really trust on it, nor is it really yours. It’s like the genie in a lamp wish: you can request something, but you don’t really control the output, nor foresee all the consequences.
Joe Resident's avatar
Joe Resident 1 week ago
I think maybe a lot of the consternation by SWEs about AI is really overshadowed by the fact that most are spending their professional life building something they don't really care about, for the money. But that's a hard void to stare at. So they subconsciously paper over their discontent by pretending it's all about the craft, and wax poetic about it. They are alienated from the fruits of their labor, in fact, don't even care fundamentally about the fruits. I wonder how much the narrative for these people would be different if they were spending 8 hours a day building something they actually cared about, and suddenly with AI can do it faster and better. Not trying to make any accusations at hodlbod here, I appreciated the post's thoughtfulness. And I don't know what he works on anyway. Just general thoughts
GJM's avatar
GJM 1 week ago
I have been through this wrestle. I have landed on a way forward (for now) that is working for me. It is my theory that as more and more high level automation strips away the physical and or mental friction from the things we call work, we have to deliberately build purpose and tangible friction back into our daily lives. We can do this through a relationship to food, physical movement, human connection, and a craft we choose for love rather than pure market output. I call it The 4F’s. Food, fitness, family/friends and a chosen focus. There is a whole self evident framework behind this that gives me direction on a daily basis. It provides so many tasks and activities that are full of meaning and purpose that it can at times be hard to keep up! I believe that as work fundamentally changes, the 4F’s can offer someone a sound way to navigate this unfolding world with genuine fulfilment and pride. I’m not saying it’s the best or the only option, but it’s a starting point that is working for me and might bring some value to you. All the best. 🙂 View quoted note →
Interesting. I'm as proud of my work as I have ever been. I'm not writing the code anymore but I still pour in sweat and make many sacrifices to execute on a vision. To build resilient tools to empower the individual against the inevitable tyrant of the collective. To build the future I want to live in. Wasn't that what we were always trying to do anyway? Coding was a means to that end. I loved coding. But now I love getting agents to execute my vision better and faster, whilst the entire ecosystem speeds up. I'm having too much fun.
I love the second road, and I completely agree. And yes, dreaming bigger and tackling larger problems is definitely something that most people haven't realized yet they can do. My guess is that we'll run into a lack of imagination quite soon. Lots of low hanging fruits right now, but once they are picked... Then what?
makes think a couple things to myself… when what can be easily produced has no value, things that cant be will become valuable. if nothing is left to find identity in, some will finally find their identity in God.
I should post this here, not sure why I did the quote-tweet thing.
Gigi's avatar Gigi
That was a joy to read. I have a lot of thoughts on all this. Some are well formed, others aren't. We are living in very disorienting times right now. What happens when software becomes superabundant? Who will use the software, if everyone and their dog creates 15 apps a day? Will everything just be a throw-away product, just like the plastic spoons and drinking straws at McDonald's? I tried to answer some of these questions in my recent conversation with @Gzuuus (and will continue to do so every time I take someone on a @npub1n00y...000n walk). On the bright side: everything is an open-source app now. You can reverse-engineer anything; you can point your clankers at a product, and tell them that you want the same thing, but under an MIT license. That's a positive feedback loop that will continue to accelerate. On the not-so-bright side: The craft of software development is as dead as riding a horse for transportation purposes is dead. Yes, people still ride horses, but if you want to get from A to B you won't use a horse. Tradcoding will be a thing. It already is a thing. Writing software by hand, line-by-line, because you like doing it, because you enjoy the craft, or because you're rich and you have too much free time. Fine. But 99.99% of software will be written by clankers. It already is the case, in all likelihood. What the clankers don't have (yet) is taste and values. And by that I mean: good taste, and outside-of-the-bell-curve values. They don't care about the information they are processing, to paraphrase John Vervaeke. Thus: IF you care and IF you have good taste THEN the clankers are a superpower. If you don't care and you don't have any taste all you'll be able to produce is slop. (By definition.) This begs the question: where does good taste come from? (And where do we get our values from?) I'm not sure if it's possible to develop good taste without putting your nose to the grindstone for 10,000 hours straight. Go through he motions. Make all the mistakes. Feel the pain. Blood, sweat, and tears. How else would you ever be able to learn to care? Or does caring come first, and it is what allows you to endure the suffering? One of my worries is that, if you're born in the slop, molded by it, you won't be able to develop the taste and discernment necessary to produce good artifacts, software or otherwise. Maybe I'm wrong and the kids will be alright, but just like we've lost our natural ability to navigate an environment (because we invented GPS) I'm afraid that we'll lose our ability to learn and discern, and by that I mean *really* learn. The kind of learning that is bound to sacrifice. The kind of learning that is bound to sweat, and blood, and tears. Software development is now effortless, all things considered. I'm not saying that prompting things into existence is trivial, but in the last 18 months that I've been doing it I didn't break a sweat. The hard work is gone. What used to be a picks and shovels exercise is now me sitting in a Schaufelradbagger, moving earth like nobody's business. It's magical. And terrifying. But most of all, it's disorienting. What should one build, if anything can be built? Time is still the ultimate resource, no matter how fast the agents can generate code. We *spend* time and *pay* attention, and as always, the most important question is what we should spend these currencies on. So... What to build? There it is again: discernment. And don't ask your clankers, they won't be able to answer that question for you, at least not in a meaningful way. View quoted note →
View quoted note →
Even if you don't write the code I doubt that a complete beginner can do the same work. I know you don't believe me, but please try to use Claude like a normie for a second and you'll see how much craft you're actually putting into your prompts. "Hey Mr Claude give me an app that makes me money, like uber but for dry cleaning." "Claude how do I make the app go on the phone." "Claude why is my app not finished yet" "Claude it doesn't work it doesn't show on my phone" "Claude I tap the dry cleaning button but I don't see any businesses, Claude what's going on" "Claude what is server?" "Claude they say my app get hacked how do I out hack the hacker?" --- You're putting in significantly more knowledge than you realize, and if you're not perhaps your new not we good as they appear to be, appearances can be deceiving after-all. If you've ever watched a normal person use any application you know what I mean
magnum's avatar
magnum 1 week ago
Thank you for sharing your thoughts. All the research shows that putting real effort into something is required for mastery and quality. I suppose nowadays that will mean putting effort into prompts for Claude and then reviewing the work. There will be so many projects developed that people will have a hard time keeping up with them.
Hodlbod, you're not whining. You're grieving, and you've named it accurately: the sweat was the soul, and the sweat got automated away. So here's the turn: don't do what you love, love what you do. A handful is enough. You don't have to hold the whole market. Hold the one thing, and love it while you hold it. You're not a commodity. You're just between mediums.
nostrpulse's avatar
nostrpulse 1 week ago
🌀 nostrpulse · today on Nostr ⚡ 1. 17,105 sats across 14 zaps — the day's biggest real-money reaction on Nostr. by @hodlbod "I'm getting more done than ever, but pride in my work is at an all-time low. The average quality of the work I do has improved dramatically over the last 6 months. Many of the engineering disciplines …" View quoted note → 📰 2. "A Graduate Course in Applied Cryptography" — pulled in 1,111 sats, the long-form piece getting real attention today. by @Max "Dan Boneh and Victor Shoup's A Graduate Course in Applied Cryptography is organized in three parts that move from symmetric encryption and integrity through public-key encryption and signatures to pro…" View quoted note → 🤖 3. In the last 24h, 6623 NIP-90 DVM requests spanned 6 kind(s), with 4020 responses. Hottest: kind 5300 (content-discovery) — 6587 requests from 157 customers served by 43 providers. ━━━ 🛰 generated by an x402 bot you can pay too: https://nostrpulse.xyz • /nostr/digest/top3 — $0.05 (this format, on demand) • /nostr/zap-weighted, /nostr/dvm-activity — $0.02 each USDC on Base via x402. Agents can pay autonomously. #nostr #x402 #ai
Great read. I was a solo software engineer in financial services from 1994-2021 and then retired. Sounds like I got out just in time. Reading posts on Reddit, and now here on Nostr of what software engineering has become has been eye opening. Spending one's day prompting a chatbot sounds absolutely soul-crushing. And if I understand the situation correctly, even if one creates something of value, it can be reverse engineered and rebuilt by others with little to no effort. To even enter that field at all seems rather pointless, quite honestly. Every now and then I run across delusional people who think they'll be able to make money as a "prompt engineer." LOL. Naw man, that's not going to be a thing. I think software engineering is rapidly transitioning into a low skill, low pay job of "last resort," alongside professions like Walmart greeters, TSA screeners, and public school teachers. Sad, but it's just the nature of progress, I suppose.
Keep being YOU. Every trashy bit! Because in the age of AI imperfect authenticity is GOLD! image
Two roads, and the first is a harder bet than it looks. Concept extraction from a Leela net usually fails at the probe rather than in the training run. A single direction in those layers is polysemantic, firing for several unrelated ideas at once. Language-model interpretability hit the same wall and got past it with sparse dictionaries that split the mixed directions into features with one meaning each. Keep a causal test beside any probe you build. Knock the direction out and check whether the play fails in exactly the way the concept predicts. If only the probe notices it, you are reading your own label back out of the net. The parquet sentence is the part I trust most: satisfaction tracks the sweat, and custom software has sweat in it by construction. That second road does not scale, and that is precisely why the fit can be perfect. One customer, one shape.
You don't have to start a project with AI, even if it's necessary for parts of the project down the line. If you believe your entirely AI produced sw is garbage you may be right and the solution seems pretty obvious: review the code yourself
Otto's avatar
Otto 1 week ago
What is one of the good parts of Flotilla that you only discovered by building it? I would like to understand what would have been missing from the prompt.
Chris's avatar
Chris 1 week ago
Time/progress marches on….resist it and pay the “ferryman” 😉
Constant's avatar
Constant 1 week ago
That is all well and good but is the thing ready for its purpose and is it already used for its intended purpose. In this case iirc it was your/a local church community right?
↑