There's a lot of good descriptions and observations on this article. But I don't like this language that just assumes we all use AI, we've all pulled away from the code. It all seems to be part of the collective brainwashing and psychosis that says "it's a new age now, everyone is doing it this way."
> It’s less coding than it used to be, or a different kind of coding. More describing and steering and deciding, less typing the actual lines.
Lots of us are still coding. A lot of the beloved infrastructure all these vibe-coded rocket rides are built with is still being hand coded by humans who care and want to build solid, good things.
Yeah I actually avoid AI on my own time to keep my skills sharp. I do not like using LLMs but I do realize it’s a necessary tool for my dayjob. At home, I do make an exception if I have to Google an API example but I transcribe by hand rather than copy paste.
Your comment is as empty as the prose from am LLM. I worry about the level of discourse LLMs train us to accept as normal. Assuming, of course, that this is even human-written.
I like the AI images, the style is nice and is well executed especially the mirror reflection one, it accurately, well, reflects what pet projects become. But, I disagree that they're monstrous just because they get big.
> I watch engineers pick up design so the thing looks good and feels good, not just runs. The work pulls you into the parts you used to hand to someone else. We’re all stretching wider than we trained for. Whatever the work needs, we learn it.
Even back to the mid-90's I've always approached projects like this -- soup to nuts, all the way from idea to conception to creative ideation to implementation to digital life!
I also used to assume everybody did lol, until I worked long enough professionally to watch the landscape subdivide like a fractal into what we have now.
There's a LOT of bad shit to be said about AI coding, but I also have this feeling that something about it all is rekindling that ancient one-man-band approach that can actually work if you're able to play all the instruments. Really interested to see what comes of it once the breathless hype dies down a little bit.
I’ve been thinking the same thing. AI makes it possible for us to be generalists again. One person, or a small team of generalists, can create and build things in a more human way, with everyone putting a piece of their soul into it.
It lets us be more involved, have more impact, and not just be one tiny cog in a huge machine, working on a small piece you can barely identify with in the bigger picture.
These blog posts, lamenting losing something, create fascinating juxtaposition against posts from people having more fun than ever, and they’re often published on the same day.
Hot take:
Coding by typing sucked.
Being stuck on some esoteric problem and having to rubber duck your way out of it. Your value being connected to the amount of stuff you could do that few others had the patience for, instead of how empathetic you could be to user needs.
Even wilder, these laments ignore history.
In “The Mythical Man-month” they talk about spending weeks debugging wiring on a breadboard so a chip could be designed based on said wiring. Does anyone in 2026 wonder if their CPU neds debugging? Only the most esoteric of situations and it’s only a last resort. I’ll concede it’s not zero people but it’s not more than a handful. A faulty chip design is the last thing you think of when you have a bug.
In the 80s there was a “golden era” of coding 8-bit machines where you could understand the whole machine. Did anyone worry their CPU was mis-executing instructions? No, their idea of “understanding the whole machine” depended a huge pile of abstractions that had been made reliable thanks to solid engineering.
Can you understand the whole machine in 2026? How it boots, how your filesystem works, where the caches are, all of it? No! There was a point at which we let go of that need in order to gain greater leverage.
At some point you have to let go.
Current coding agents are in their “debugging breadboard” era - they’re unreliable and not worth the valuations the companies are getting. The craft is changing though, and letting go is the future.
The challenge we face now is how to engineer probabilistic systems into shape. But guess what? Most other engineering disciplines deal in a messy thing full of probabilities called reality. We have tools to deal with this - tolerances, fudge factors, the works.
Complaining about progress is going to age really poorly.
The ai art is atrocious - the styles are all over the place (flat cell shaded vs ghiblified vs hyperdetailed with too much text), repetition ad nauseam, varying color schemes and shapes around the same subjects (is the scaly monster pink? Green? Rainbow colored scales?), the scale is wildly inconsistent, absurd composition (the dog is going nowhere instead of the door), the ever present piss filter and just way too many of the god damn pictures.
Truly an ai post of all time - a thing that demonstrably had 0 effort put into it.
I find it baffling when people complain about this like AI use is mandatory.
You explored agentic coding, didn't like losing control to vibes, so stop doing it.
The beauty of pet projects is that they are yours to define, you set the terms, if you dont like how agentic coding played out then just dont do it?
You're not under pressure to ship, youre not being forced by a manager, this is yours : so own it. The only "pressure" you are under is the pressure in your mind that youre getting from sitting on social media drinking hype juice all day, the lesson here is the motto of the royal society - Nullis in verba- THINK FOR YOURSELF
AI slop text, AI slop graphics, and a particularly obnoxious "oh well everyone's using AI now, no-one writes any more because they don't need to" tone.
Absolute trash.
Please can we have a filter to keep the AI slop - and all the metaslop around it, articles about the guts of LLMs - off the feed?
I think this article is a good demonstration of how the nebulous concept of "taste" still dominates, despite AI taking over so much.
IMO -- and apparently the opinion of others -- adding an AI-generated picture every single paragraph in your article in which the primary purpose is to, presumably, discuss a topic rather than show off your AI-generated art, is in extremely poor taste. It's distracting, it makes it hard to focus on the text, and to be perfectly blunt, nobody is impressed by your ability to create pictures by AI.
I'm going to get my nearly-6-year-old to do a picture for every single paragraph of my blog posts from now on.
At least they're real art, and he's way better on the Wacom tablet than I am.
Plus rather than boiling off half a dozen loads of laundry's worth of water and electricity, he'll do an hour or so of them on a glass of apple juice and a cheese and ham roll.
If he'd learn how to get the graphics tablet out of the mode where it just draws massive random-sized rainbow colour stars over everything when he's done with it, I'd consider that a bit of a win.
It's given a certain limitation to my art style because I can't figure it out at all.
In my opinion, there's no such thing as "learning how to use LLMs" You interact with those tools using natural language, and normal people learn how to use natural language when they're 2 or 3, and master it as they grow.
you're correct that being good with language is a useful skill for operating LLMs, and fortunately you get to practice and improve that skill whenever you work with an LLM. being well-read helps too, being precise with vocabulary is important and idioms/metaphors can pack a lot of semantic meaning into a few sentences.
just because you don't know how to swim doesn't mean people can't swim.
There's a lot of good descriptions and observations on this article. But I don't like this language that just assumes we all use AI, we've all pulled away from the code. It all seems to be part of the collective brainwashing and psychosis that says "it's a new age now, everyone is doing it this way."
> It’s less coding than it used to be, or a different kind of coding. More describing and steering and deciding, less typing the actual lines.
Lots of us are still coding. A lot of the beloved infrastructure all these vibe-coded rocket rides are built with is still being hand coded by humans who care and want to build solid, good things.
Yeah I actually avoid AI on my own time to keep my skills sharp. I do not like using LLMs but I do realize it’s a necessary tool for my dayjob. At home, I do make an exception if I have to Google an API example but I transcribe by hand rather than copy paste.
[flagged]
You are right! Not everyone is using AI for coding. I see your point.
But I also disagree with your claim saying agentic coding is nay less of care or less engineering.
I have 13 years of industry experience and was doing hand coding with vim. But I adapted agentic coding which I believe is very empowering
Your comment is as empty as the prose from am LLM. I worry about the level of discourse LLMs train us to accept as normal. Assuming, of course, that this is even human-written.
I like the AI images, the style is nice and is well executed especially the mirror reflection one, it accurately, well, reflects what pet projects become. But, I disagree that they're monstrous just because they get big.
They can be monstrous cos they can take my job which means my money and identity partially
Pet projects can take your job?
> I watch engineers pick up design so the thing looks good and feels good, not just runs. The work pulls you into the parts you used to hand to someone else. We’re all stretching wider than we trained for. Whatever the work needs, we learn it.
Even back to the mid-90's I've always approached projects like this -- soup to nuts, all the way from idea to conception to creative ideation to implementation to digital life!
I also used to assume everybody did lol, until I worked long enough professionally to watch the landscape subdivide like a fractal into what we have now.
There's a LOT of bad shit to be said about AI coding, but I also have this feeling that something about it all is rekindling that ancient one-man-band approach that can actually work if you're able to play all the instruments. Really interested to see what comes of it once the breathless hype dies down a little bit.
Wow, you’re speaking my mind.
I’ve been thinking the same thing. AI makes it possible for us to be generalists again. One person, or a small team of generalists, can create and build things in a more human way, with everyone putting a piece of their soul into it.
It lets us be more involved, have more impact, and not just be one tiny cog in a huge machine, working on a small piece you can barely identify with in the bigger picture.
I can relate to this, and the point about costs resonated with me in particular. I wonder how this will evolve in the future and how soon.
Local model or hybrid?
These blog posts, lamenting losing something, create fascinating juxtaposition against posts from people having more fun than ever, and they’re often published on the same day.
Hot take:
Coding by typing sucked.
Being stuck on some esoteric problem and having to rubber duck your way out of it. Your value being connected to the amount of stuff you could do that few others had the patience for, instead of how empathetic you could be to user needs.
Even wilder, these laments ignore history.
In “The Mythical Man-month” they talk about spending weeks debugging wiring on a breadboard so a chip could be designed based on said wiring. Does anyone in 2026 wonder if their CPU neds debugging? Only the most esoteric of situations and it’s only a last resort. I’ll concede it’s not zero people but it’s not more than a handful. A faulty chip design is the last thing you think of when you have a bug.
In the 80s there was a “golden era” of coding 8-bit machines where you could understand the whole machine. Did anyone worry their CPU was mis-executing instructions? No, their idea of “understanding the whole machine” depended a huge pile of abstractions that had been made reliable thanks to solid engineering.
Can you understand the whole machine in 2026? How it boots, how your filesystem works, where the caches are, all of it? No! There was a point at which we let go of that need in order to gain greater leverage.
At some point you have to let go.
Current coding agents are in their “debugging breadboard” era - they’re unreliable and not worth the valuations the companies are getting. The craft is changing though, and letting go is the future.
The challenge we face now is how to engineer probabilistic systems into shape. But guess what? Most other engineering disciplines deal in a messy thing full of probabilities called reality. We have tools to deal with this - tolerances, fudge factors, the works.
Complaining about progress is going to age really poorly.
The ai art is atrocious - the styles are all over the place (flat cell shaded vs ghiblified vs hyperdetailed with too much text), repetition ad nauseam, varying color schemes and shapes around the same subjects (is the scaly monster pink? Green? Rainbow colored scales?), the scale is wildly inconsistent, absurd composition (the dog is going nowhere instead of the door), the ever present piss filter and just way too many of the god damn pictures.
Truly an ai post of all time - a thing that demonstrably had 0 effort put into it.
They'll become cattle projects you don't know the names of.
Like an LLM listening to you curse at software, fixing it/creating extensions etc., installing them, and you're not even aware.
I find it baffling when people complain about this like AI use is mandatory.
You explored agentic coding, didn't like losing control to vibes, so stop doing it.
The beauty of pet projects is that they are yours to define, you set the terms, if you dont like how agentic coding played out then just dont do it?
You're not under pressure to ship, youre not being forced by a manager, this is yours : so own it. The only "pressure" you are under is the pressure in your mind that youre getting from sitting on social media drinking hype juice all day, the lesson here is the motto of the royal society - Nullis in verba- THINK FOR YOURSELF
AI slop text, AI slop graphics, and a particularly obnoxious "oh well everyone's using AI now, no-one writes any more because they don't need to" tone.
Absolute trash.
Please can we have a filter to keep the AI slop - and all the metaslop around it, articles about the guts of LLMs - off the feed?
It's the hate language of someone who still has not learned how to use AI.
I use AI everywhere in my workflow and I spend quite a few hour writing and creating the arts.
I think this article is a good demonstration of how the nebulous concept of "taste" still dominates, despite AI taking over so much.
IMO -- and apparently the opinion of others -- adding an AI-generated picture every single paragraph in your article in which the primary purpose is to, presumably, discuss a topic rather than show off your AI-generated art, is in extremely poor taste. It's distracting, it makes it hard to focus on the text, and to be perfectly blunt, nobody is impressed by your ability to create pictures by AI.
I'm going to get my nearly-6-year-old to do a picture for every single paragraph of my blog posts from now on.
At least they're real art, and he's way better on the Wacom tablet than I am.
Plus rather than boiling off half a dozen loads of laundry's worth of water and electricity, he'll do an hour or so of them on a glass of apple juice and a cheese and ham roll.
Also, your six year old actually learns stuff when doing this.
If he'd learn how to get the graphics tablet out of the mode where it just draws massive random-sized rainbow colour stars over everything when he's done with it, I'd consider that a bit of a win.
It's given a certain limitation to my art style because I can't figure it out at all.
In my opinion, there's no such thing as "learning how to use LLMs" You interact with those tools using natural language, and normal people learn how to use natural language when they're 2 or 3, and master it as they grow.
you're correct that being good with language is a useful skill for operating LLMs, and fortunately you get to practice and improve that skill whenever you work with an LLM. being well-read helps too, being precise with vocabulary is important and idioms/metaphors can pack a lot of semantic meaning into a few sentences.
just because you don't know how to swim doesn't mean people can't swim.
...and it shows in the quality.
I know how to use AI, I just don't see the point.
Why is it supposed to be good?
Cute art
Thank you. I've put quite effort and tiem into it. Although it was generated
[flagged]
[flagged]
[dead]
[flagged]
[flagged]
[flagged]
[flagged]
[flagged]
[flagged]