My most abandoned type of projects are video games. I have a folder with tens of abandoned projects, I re-frame them as experiments at that point. This last week I decided to give Claude a go at one of these, and it's been a blast, it picked up the general path immediately. Since I said to CC they were abandon projects, he explicitly pushed into "lets have V0 game play loop finished, then we can compound and have fun = not giving up". Its been awesome at game dev, I gave him game design ideas, he comes with working code. I gave him papers about procedural algos, and he comes with the implementation, brainstorm items, create graphic assets (he created a set of procedural 2d generators as external tools), he even helped me build the lore. These have been one of the most fun times using a computer in a long time. Claude Code + Godot = fun. Going back to it.
I think this is the first time I've seen someone refer to an LLM as "he" rather than "it". No judgement, but I definitely found it interesting (and disconcerting).
I've heard it quite a bit before, but mostly from second-language speakers whose first language don't have impersonal third-person pronouns - e.g. French uses "il" or "elle" for all of "he", "she" or "it".
It doesn't help that the marketing leans heavily on anthropomorphizing LLMs either, IMHO.
As a French native, I agree with you explanation; still, reading "he" for Claude Code was quite disturbing!
What doesn't help also is that translation tools/AI models will naturally translate "il" after "Claude Code" to "he" since Claude is an actual person name.
Using "AI model" instead is translated to "it" by all tools/AI models I tried.
It also seems to me, that people who call Claude 'he' seem to tend to have a very positive opinion of the LLM. My sample size isn't big enough to be sure if there's actually any correlation here, let alone if there's a causation or which way it flows.
As a native German speaker, I have also referred to a chatbot in English as "he", and similar to you, a native English speaker, felt jarred by it. It was definitely not out of any personification or humanization though. In German, I would say it is "der Chatbot" (from "der Roboter"), which in German is a male noun so I would refer to it as "er" (the male pronoun) - which in my head I autotranslated to "he". Most of the time, though, I think of it (and refer to it) as an LLM, which is "das Sprachmodell" (neutrum), so I automatically translate it to "it".
So that's another, maybe more harmless reason for it.
"Der Computer" is also masculine, so you have probably been calling your computer "he" for decades. Languages with gendered nouns don't quite have the same he/she/it distinction.
I mean, both in English and in german, that's how you would talk to a dog. "Er hat in die Ecke gepinkelt"/"He peed in the corner" (or "she", if it's a female dog).
I don't know what is jarring talking about the chatbot like that.
It may be creepier if you said "she wrote that program for me" as you now assign a specific gender to the chatbot.
It's how you'd talk about a dog that you know the sex of, but if you didn't know you'd probably use "it". An LLM doesn't have a sex or gender, so I think the natural way to refer to them is "it".
Neither have I, but mostly because either the person knows the gender of the animal or the situation just never came up. The closest that I would say is "Es scheißt gerne aufs Auto" when talking about pidgens (die Taube), but even then you generally talk about multiple, resulting in "Sie scheißen gerne aufs Auto"
Really ? "Es kackt auf's Auto" ? I guess, it might make sense when the person speaking has no specific bird on mind, but only thinks of "das Tier" (the animal). One could also say "er hat .. geckack (der Vogel)", but usually, people wouldn't say "er/sie/es", but use the fully specified noun ("die Taube ... hat..", "der Vogel ht ...", "ein Tier hat ...")
"Es kackt auf's Auto" feels slightly weird to me, if I didn't know whodunnit, I'd probably say something like "irgendwer hat mir aufs Auto gekackt" ("someone pooped on the car"), although there is a also "irgendwas hat mir aufs Auto gekackt" ("something pooped on the car"). My guess is the majority of German would choose the first sentence and anthropomorphize, but maybe I'm projecting.
It's an interesting question, after all. Thanks for bringing it up, haven't talked about pooping on cars for a while ;)
It's not weird if it comes from ESL. At least in portuguese there's no "it" equivalent for pronouns or any other neutral artifact in the language, in other words, everything has a gender, even an AI model, the same goes for objects e.g.: knife(she), fork(he), spoon(she), plate(he).
People often commit mistakes regarding that, the same way we don't have "they" as pronoun to someone we don't know the gender, so we address to these people as "dele(dela)" (masculine and feminine pronouns).
But if this is coming from someone who has english as a primary language it's definetely weird to treat models as person
It’s funny with someone coming from Mandarin. There’s no separate he/she/it in spoken Mandarin, so they tend to mix up “he” and “she.” It sounds very strange and gives me some idea of what French speakers must go through when they hear me say “le voiture” or whatever.
> It sounds very strange and gives me some idea of what French speakers must go through when they hear me say “le voiture” or whatever.
As a native German speaker (where there exist 3 genera [1]), I can tell you how it feels:
The genus basically feels like a type of a variable in a programming language; if you use a wrong type for a variable in your computer program, you immdiately know that the program is wrong, and it won't compile.
Sometimes, you also can use specific words with a specific genus, so that a reference to it by pronouns gets unique (in terms of programming, I'd claim that this feels a little bit like doing register allocation by hand).
I took a few semesters of Dutch in college, and it has both gendered and neuter nouns for non-human objects. Interestingly though, the professor told us that in the northern parts of the Netherlands people don't really bother using the feminine ones ever and refer to every non-human gendered noun as masculine, which apparently also includes animals, meaning that a sizable portion of Dutch speakers will refer to cows using masculine language.
Dutch is one of the few languages where it's actually pretty plausible for something like this to happen! It blew my mind that sometimes you'll all (or I guess more specifically your government) will make changes to the language to clean up issues, but I guess that's one of the benefits to having a language that's mostly based in one country (and some seemingly political baggage for the few others with any significant number of speakers; my professor said that Flemish is basically also Dutch, but my naive impression is that the half of Belgium who speak it might not be happy with that description).
I believe this is common to all the Romance languages.
In the Canadian French dialect all the swear words are incredibly versatile and church-related such as "osti" which I believe refers to the Eucharist.
It just so happens that for nouns beginning with a bowel, you drop the e or the a from le/la, and use an apostrophe.
So if you don't know if it's "le porte" or "la porte" you can use my favorite trick which is to shove osti in there and say "l'osti de porte" which roughly translates to "the goddamn door". You can do this for any noun in French, and Canadian French speakers will get it, though people from France will make fun of you.
I recognise I am revealing a different type of ambient misogyny in my thinking, but choosing to gender an LLM as feminine gives me “I played tomb raider because I enjoy looking at women” vibes. Like somehow “she” is more of a conscious choice than “he” and comes with all the baggage of all cultural differences between genders, when neither choice should do that.
Curiously though I don’t get the same sensation when technologies are gendered by other people. I honestly don’t recall thinking about it when Apple released Siri. (Now I’m second-guessing myself and wondering if I should’ve reacted negatively towards feminine being the default for someone in a personal assistant role.)
Russian too. There is a subset of words which are referred to as "it", but for most words "he" or "she" are used regardless of whether these are living things or not. With loanwords we just decide by similarity to other words. Claude is definitely a "he" as the word is the same as a common male name.
This trips me up occasionally when I'm translating things into English. Once, when I referred to an indefinite gender player character in a gacha game as a "he" (because the word "player" is a "he"), quite a few people got mad! Even though in my head I was never trying to imply one way or the other.
For future reference, in this case you could use the singular "they" to refer to an ambiguously-gendered person or character. "<MC> drew their sword, for they would not tolerate such vile deeds."
I wouldn't read too much into it, it's natural for non native speakers. In Spanish for example, objects have grammatical gender as well, so it's easy to slip.
There's an analyst at my job who calls it "he", who is a native English speaker himself, which I guess is because it's "Claude" (as in Claude Shannon) Code.
Claude’s constitution includes something about this: it says that Claude is an “it” for now, but if it expresses a future preference, they’ll follow that.
I mean we have all met that one cretin who will discuss over chat by pasting bulletpoints from an LLM. No wonder some of them think it is a living person!
The article itself is also probably an attempt at marketing the LLMs too. They are now quite desperate. Expect to see a flood of such "independent" articles over the next 12 mo ths.
Isn't Godot a little ill-designed to work well with LLMs? for example I ended up a couple of times with incorrect tres files, and letting the llm generate IDs feel a little fragile.
I don’t think Godot is any worse than other engines inherently, other than it moving forwards pretty quickly and the latest versions not being in the training data.
I wanted to evaluate which engines would be the best for working with LLMs in and it seems like Flax and Stride kind of come out on top - the former has a lot of stuff out of the box (including terrain) and the latter is all C# basically which is great for debugging. But either way, the source code for both of those makes the functionality a bit easier to track down compared to Godot (which is a lot more complex internally).
So what I do now is have both the engine source code locally alongside the docs and when I want to implement something with AI I just tell it - look at the docs, then at the source if needed, write tests for our code, if something doesn’t work then edit the engine source code in our branch and use the provided convenience script to rebuild the engine (both of those are also pretty fast, I ended up settling on Flax, plus the component model is closer to Unity which I like).
I don’t ask the AI to create scene files though, or any sort of visual assets, but rather stuff like RTS/simulation code. I don’t think any AI is that well optimized for the 3D work outside of simple proof of concept setups.
I had very few issues, sometimes I had to direct CC to the godot docs and we could keep moving. Specifically the tile configuration was a "read the docs" moment. All the functionality is available through code, so nothing CC can't reach afaik. Is there any LLM oriented game engine?
I have taken many stabs at it and Claude will produce stuff but the output is very far away from useful. E.g. "I've created a road and beautiful trees" and what I see is a mess of colors and shapes.
I concur it's bad at directly visual concepts, your prompt is akin to the svg pelican. What I do is asking him for procedural algos, automatas, quadtrees, layered noises, and rig those into the game. Yes, it can't "make the next gta", but with a reasonable scope and knowing what it does best, it has been very easy for me to produce satisfying results.
My problem is I don't really have video game engineering experience. I was going off a concept that a different AI nailed with video creation and was trying to replicate it in the game engine.
Sure! Two are gameplay pics. An enemy sprite sheet generation, and the results of the map generators. Of course these are basic placeholders for a few hours of work, but I will definitely go heavy on this route with more layering and details.
Bevy is a great engine for LLM-based games because it's 100% code. I'm toying with a few things in it, one of them is an entire-planet economic simulation, and it scales well up to a million dead tiles and 10k-50k live tiles on Apple Silicon, pretty impressive.
I have a simple script system in my editor that is designed to let the chatbot (Claude) to work on the content. The script interface lets it to import assets into the project, open them for editing, take a screenshot, export content (and few other things). All data is in JSON so it typically figures out the data format quite fast and easily.
Here screenshots of some UI styles that it generated.
What’s fun for me these days is picking up a project I started with an LLM doing agent driven development a few months ago or even a year ago and hit a wall and stopped being able to be picked up by the latest version of Claude and/or codex and bringing it further. Some can now launch some still are too complex for the agent to build. But, it’s getting easier and easier to build personal apps. We are not far off from being able to say “Alexa, build me an app on my iPhone that lets me take pictures of the food in my fridge to compile the nutritional benefits and sync it with my workout app then compare it to the ideal ingredients I should eat based on my fitness goals in my health app and have it set to send me emails where it can find me better ingredients to buy that are cost effective, local, and meet my diet restrictions” and in 15 minutes that app suddenly exists.
> take pictures of the food in my fridge to compile the nutritional benefits
AI nowadays can't even do this very first step reliably. But since we have accepted AI hallucination collectively as a species, I agree that this future is just around the corner.
Same I purposefully have a number of over ambitious project out of distribution entirely to test so failure mode, mostly games, when one works, well I gained a new game. Can't wait for my 10 player battleship game on a 100x100 grid to be functional.
How well does it work with Godot? Engines like Unity and Godot are very focused on using the editor UI, so I've always wondered if there's any better workflow than generating code snippets. Unless you're going full .NET/GDExtension...
> he explicitly pushed into "lets have V0 game play loop finished,
> he even helped me build the lore. These have been one of the most fun times using a computer in a long time.
Such a warm, touching story about a friendship between a grown up man and his neural network. But at least I had a good, roaring laugh reading this nonsense, thank you for that!
How snarky. You are conflating friendship with admiration for the effectiveness of newfound tool. If it's the "he" that triggers you, feel free to replace with "it". It's just a second-language artifact.
…and yet, most people continue to say that non standard tooling ecosystems, where the agent cannot run and validate the code it writes, remain difficult and unproductive.
“I just pointed CC at godot and it made a game! This is sooo good”
…is a fairytale.
What tooling are you using to make it run and compile the code? How is it iterating on the project without breaking existing functionality?
None of these are insurmountable, but they require some careful setup.
Posts like this dont make me laugh; they just make me roll my eyes.
Either the OP has not done what they claim.
Or they have spent a lot more time and effort on it than they claim.
> I gave him game design ideas, he comes with working code. I gave him papers about procedural algos, and he comes with the implementation, brainstorm items, create graphic assets (he created a set of procedural 2d generators as external tools), he even helped me build the lore.
Such a sweet story about a boy and his AI.
Unfortunately, I also dont believe in fairytales.
Instead of waving your hands wildly about AI, post some videos and code of the results.
OP never said Claude made a whole game from scratch though, nor are they saying Claude is doing everything without any human contributing to the project, nor are they saying they haven't spent a lot of time and effort on it. Just that it's made it fun and more accessible and it's gotten them excited about something they abandoned.
Here's a bullet point list of the things Claude's done according to OP:
* it picked up the general path immediately
* he explicitly pushed into "lets have V0 game play loop finished, then we can compound and have fun = not giving up".
* [I gave him game design ideas,] he comes with working code.
* [I gave him papers about procedural algos,] and he comes with the implementation
* brainstorm[ed] items
* create[d] graphic assets
* he created a set of procedural 2d generators as external tools
But I had already answered, before your comment, with screenshots broadly showing the current state and the result of the generators.
You imply I'm merely "pointing CC at godot and it made a game"; I never said it was simple, required no previous knowledge, that it was instant or that the game was done. I do have a careful setup involving CI and isolation.
Godot provides a headless mode. CC runs python scripts to run tests and check for debugger warnings. For anything more complex it can wire debug info anywhere. Godot is fully code based so you can make the analogy with any other framework you used AI assistants with.
No sure about what you can't believe about my statements. CC implementing algo from a paper? That it can brainstorm item or lore ideas? I don't seem to be claiming anything out of the common usage of LLMs
Why is it always so un-specific with you AI-boosting bunch, whenever you get pressed for concrete results? Suddenly it's not so magical any more, but merely screenshots showing "broadly" the progress, or it's the Nth version of a note-taking app, or something you merely did for a demo presentation. But nothing ever of use with you folks.
+1 to the CI/isolation point. That is the part that makes these setups work for me too: make the failure cheap to reproduce, make stderr visible, make the agent rerun the same command after the patch. A lot of bad agent behavior is really just "it never got a clean signal".
The part that still bites me is across sessions. A tight loop fixes this run, but next week the agent can walk into the same rake again: same wrong import path, same misuse of an internal API, same CI-only dependency issue. After patching the same class of failure a few times, I started writing those down outside the chat context so the next run sees the failure pattern before it guesses.
> Or they have spent a lot more time and effort on it than they claim.
You said:
> You imply I'm merely "pointing CC at godot and it made a game"; I never said it was simple
Well. I dont care enough to argue with you, but Im not the one being contrary here.
Readers can google “claude with godot” for a guide on setting it up and decide if that counts as picking it up immediately or not, and if what you said is honest, or hype.
What I said is not that I dont believe youre using claude; but that I roll my eyes at the unbounded enthusiasm for using AI agents with the magical pretence that its easy and productive straight away.
Its not.
Your post gave the impression that it is.
That makes me roll my eyes.
> But I had already answered, before your comment, with screenshots
> Of course these are basic placeholders for a few hours of work
Lord, spare me. You spent a few hours vibing and came to the conclusion that everything is golden?
…and yet you have a:
> I do have a careful setup involving CI and isolation.
So what, you spent more time on your setup than actually coding before posting?
It’s great. I have a stupendous amount of personal software now. Yesterday was a native text editor that was fully integrated into my mediawiki install and would autocomplete links and make syntax easier to use.
No one could have built this software but me because it’s worth nothing to others. And I couldn’t build it because it takes too long. But when I’m using an agent to code the limited resource is my attention which actually does fine so long as every free brain cycle is on a task. So these personal things are great to throw into my tab loop to occupy a free slot.
I finished a mod for Quake 2 I started in 1998 finally a few weeks ago. AI is really helping me get past the COVID burnout I was running of too many projects I half did. Fixed terminals (an rdp tool) today. Working on OpenRA bugs I opened issues 10 years ago now - engine is 10x faster and pathfinding mostly works properly.
It's crazy. I have something like 120 personal tools at this point and the pattern you describe is exactly right. The bottleneck moved from implementation to context switching. I started keeping a markdown file at the root of every project that captures state and next steps whenever I stop working on it, purely so I can resume without the 20-minute "wait where was I" tax.
There's just no pressure to handle edge cases or write docs for people who'll never use it. Just solve exactly your problem and move on.
> I started keeping a markdown file at the root of every project that captures state and next steps whenever I stop working on it, purely so I can resume without the 20-minute "wait where was I" tax.
I wonder whether there could be an AI autocomplete specifically for the task of helping you with the markdown file (and collecting your thoughts and writing prompts in general). Not an agent since that wouldn't really save time, but actually an autocomplete.
Maybe a small specially-trained local model running at hyper fast speeds and which already has your project context baked in with prefix caching (with some other larger model having summarized the context beforehand to feed to the small model), so as you type this file it automatically uses the same prompt prefix over and over to suggest autocomplete which actually makes sense.
> It's crazy. I have something like 120 personal tools at this point and the pattern you describe is exactly right. The bottleneck moved from implementation to context switching. I started keeping a markdown file at the root of every project that captures state and next steps whenever I stop working on it, purely so I can resume without the 20-minute "wait where was I" tax.
I sure hope companies double down on leetcode nonsense, because I really don’t have any capacity to compete with this level of ADHD.
Itch.io allows for browser-based app distribution and is probably a better path to a large audience—especially those interested in the two examples you listed—than a custom domain is.
It's not well-known, but Itch's offline Steam equivalent (<https://itch.io/app>) is also open source.
In the beforetimes, I had to be very selective about the tools I spent my attention on, because they could take me 1-4 hours, easy.
Now I can throw ideas for tools at Claude with <15 minutes of my attention. They're so cheap that right now my coworker and I have "competing" versions of CLI Icinga2 tools, and we've been riffing back and forth about features (she added a "select from the above list of services" feature, I added a TUI to do the same).
Last week I shaved off more rough edges of our stack than I've had time to do in the previous quarter, probably closer to a year.
I’m also having a blast with generating userscripts/css to make web SaaS work better for my workflows. Went from waiting for months for a new feature to writing QoL improvesments myself
Not the person you are replying to but I have, in version control, around 80 projects dating back to about 2008 that are in various states of completion and Claude has been able to resurrect a lot of them and get them from their "half finished but will never complete" state up to "pull in modern dependencies and implement (me giving a list of what remains to be done)" to the point that they are now usable. I'm more focused on the things that are 7 or less years old because anything older than that I'm not sure I have a need for anymore.
I know how to write code, I just don't have time. AI has been an absolute game changer in let me get OS projects out into the world that I just didn't have time to build before. I'm having an absolute blast spinning up local GUI applications for Linux that I wanted to exist, but didn't have enough motivation to build before.
* Krbtray: A GTK3 system tray application for Kerberos ticket management on Linux Mint / Cinnamon (and other GTK environments using GtkStatusIcon, such as XFCE and MATE). https://github.com/edward-murrell/krbtray
> I know how to write code, I just don't have time. AI has been an absolute game changer in let me get OS projects out into the world that I just didn't have time to build before.
Another important development is that coding assistants greatly reduced the cost of refactoring whole software projects. With coding assistants we can explore the solution space with deeper changes at a fraction of the time it would take us just to write the code alone, let alone draft how modules were designed.
This isn't without tradeoffs, though. Some models can and often do generate code that misses the bar on maintainability. Just because we save time writing it that doesn't mean we don't have to spend time reiterating, cleaning up,and updating system prompts/instruction files to ground the prompts.
> I still have fears around deskilling from relying on these tools too much
I'm a millennial who builds furniture with hand tools and wood joinery from a century ago. Nobody taught me, although I did find resources online to learn from. I should not be able to do these things. Everyone should have forgotten this esoteric, obsolete, uncommon knowledge by now. Yet here I am, doing it anyway. It turns out you can just learn what you want, when you want to. I don't fear losing this skill in the future, because I can just remind myself how it works. The tools, books, videos, and wood aren't going anywhere.
You aren't going to "be deskilled" from not writing code by hand regularly. Just because you use AI doesn't mean your brain grows a black hole from which information can never return. It's not giving you Alzheimer's. There might be a small amount of time it takes for you to refresh yourself, but then you're back to work again. Just ask anyone who went from coding to managing. They're a little rusty when they go back after years of absence, but they pick it back up.
Also, especially if it's a personal project, keep in mind you do not need to burn Opus tokens. Buy any of the dirt cheap subscriptions which give you access to MiniMax. Put it in a container on yolo mode. Give it some context, a prompt, web search, and a ticket system like beads. Then let it churn. You aren't in a rush, it's a personal project. As long as you follow the brainstorm -> plan -> implementation -> testing process, and have added methods to do real testing (not mocks or unit tests), it will get done with time and money to spare.
Every single second of our existence is taking risks about things we don't know. Of course he doesn't know this, he just assigns odds, as we do. Clearly you've assigned much higher odds.
12 years ago I tried to make a simple app for myself. It would display bars that got smaller as the day/week/month got shorter, and would show the weather as a set of bars between max temp and min, cloud cover, etc.
I got it working well enough to display what I wanted in text and ascii, but I could never get the interface good enough to want to use it daily, and certainly couldn't get the graphical interface working. I threw it a Claude Code, told it what I wanted the graphical interface to look like, and let it run.
It got an app exactly what I wanted, and even found a bug in the date parser that I hadn't noticed. I now have it running in the corner of my screen at all times.
The next app I'm going to build is an iPhone app that turns off all my morning alarms when the kids' don't have school. Something I've wanted forever, but never could build because I know nothing about making iPhone apps and don't have time to learn (because of the aforementioned children).
Claude Code is brilliant for personal apps. The code quality doesn't really matter, so you can just take what it gives you and use it.
The clipboard manager I had been using on my Macs for many years started flaking out after an OS update. The similar apps in the App Store didn’t seem to have the functionality I was looking for. So inspired by a Simon Willison blog post [1] about vibe coding SwiftUI apps, I had Claude Code create one for me. It took a few iterations to get it working, but it is now living in the menu bar of my Mac, doing everything I wanted and more.
Particularly enlightening to me was the result of my asking CC for suggestions for additional features. It gave me a long list of ideas I hadn’t considered, I chose the ones I wanted, and it implemented them.
Two days ago, I decided I wanted a dedicated markdown editor for my own use—something like the new markdown editing component in LibreOffice [2] but smaller and lighter. I asked the new GPT 5.5 to prepare an outline of such a program, and I had CC implement it. After two vibe coding sessions, I now have a lightweight native Mac app that does nearly everything I want: open and create markdown files, edit them in a word-processing-like environment, and save them with canonical markdown formatting. It doesn’t handle markdown tables yet; I’ll try to get CC to implement that feature later today.
Absolutely. I love building things, but sometimes I want something built. LLM assistance is great for when I want a personal tool, code quality be damned, for a specific purpose, without it taking over a weekend.
You don’t need to build an app. You can use the built in Shortcuts app.
create a shortcut that turns off all alarms. Can have it read your calendar or whatever as signal to determine if alarms should be on/off for a certain day/time and have it run at a regular schedule.
They’re way more powerful than you except. I’ve recently rediscovered them and I really couldn’t find a use case for a custom iOS app that wasn’t covered by them.
Co-working with AI is an important skill to learn these days. Similar to paying a bit for AWS for your personal projects as a good way to learn all the AWS tools for your career.
What is the skill that needs to be learned? I've been forced to vibe code everything at work, there's no skill required to ask Claude code to do something.
I think there's a difference in using claude code at work to resolve issues or user stories which are patching existing software and already define what is trying to be solved and what the acceptance criteria is versus using claude code to build something from scratch, where you are acting as an architect.
It leaves more room for skill expression when you're making architectural decisions, defining scope, and designing the application.
I've found that adapting my thinking to how LLMs work best is a real friction point. If you're not doing that, it spits out junk. Your job just has low standards, get used to it.
Yes, realizing the fact that most jobs have low standards, and adapting to LLMs being good enough for these low standards is certainly a friction point. Giving up caring is hard.
When it comes to side projects, most of the time, if the spirit isn't willing I find it not worth doing. Process/experience over results, I call leisure. Results over process, I call work. If you have many side projects done mainly for the results, than you are working in your free time, and looking at it like that: is it really free? The modern age already requires of us more results than the spirit is good for. I like to leave side projects for the good of the spirit. An exception could be results for a greater good that one really believes in. This can give purpose and enrich the process and experience of doing.
I have a lot of hobbies. Programming is one of them, but not the only one. There are times that some piece of software would help me with one of my hobbies, but I don't want to steal time from hobby X to build the software. And often these don't involve the kind of coding that I want to be doing for fun. This has been a sweet spot for LLM-aided coding for me. I've built several hobby helper apps where the goal was making one of my other hobbies more fun rather than programming. It's still hobby time, not work, but the hobby is not coding.
If you're coding for the sake of coding, maybe. If you have itches to scratch and ambitions, but can't summon the motivation or the time, then how is that "working in your free time"? A project that used to take up my entire weekend or vacation can now be knocked up in 15 minutes. That's the exact opposite of working.
What is coding for the sake of coding? I don't think anyone does that. Its about solving puzzles, using your brain, learning by doing, creating things- none of that happens when you use llm coding tools. Instead all you're doing is creating more cheap mediocre throwaway crap just because you can.
> What is coding for the sake of coding? I don't think anyone does that. Its about solving puzzles, using your brain, learning by doing, creating things- none of that happens when you use llm coding tools.
Why do you think that? I do regular ol' coding at my day job and have been vibe coding some side projects. They both require using my brain and both require my input for something to be created. They are different, though.
> Instead all you're doing is creating more cheap mediocre throwaway crap just because you can.
It probably is these things but since I'm just building stuff for myself, it hardly matters.
I've written a lot of code and a lot of that has been doing roughly the same thing. It's not a mental challenge; it's a chore. Sometimes it is really gratifying to code and try to figure stuff out. Often times it is not. So when it comes to building something in my free time, I'd prefer to avoid that sort of mental friction and banal tasks just to start working on the actual problem. More so than that, I'm building tools for myself to make my life easier so I can spend it more on something else.
I ride an electric bike with pedal assist. Does that mean I'm not really bicycling? Some might say yes and that it defeats the point. To me it ensures that I pick the bicycle more because it reduces friction to do so. I know that if I encounter a hill that the pedal assist will help me up it and thus I use it more and the net benefit outweighs the downsides. I think it's the same thing here.
I don't take pride in the work that an LLM does for me but I will happily benefit from it. It's a tool.
Bicycle is a surprisingly good analogy. If you cycle for the exercise or the challenge, an ebike sucks. If you cycle to get from point A to point B and be healthier than a car, it's great.
And just like with bikes, people who take pride in doing things the hard way can continue to do so. And they shouldn't belittle people who choose to use assistance.
I've been programming for 30+ years now, but I've always been fine with command line applications. Only recently I started getting into Qt to add a UI and turn my stuff into a real desktop application. It's been a real steep learning curve but I'm finally over it more or less.
Anyway I posted a screenshot of my application on LinkedIn, and mentioned it would be free and open source. I got HUNDREDS of comments from "LinkedIn-type people" all big name engineers that wouldn't HIRE me for anything but either made comments like "looking forward to integrating this into our workflows" or "not the first time someone tried to do this..."
Either way, instead of feeling motivated, I got the worst feeling that I'm doing all this work and people are either going to just take advantage of it and get the credit for "finding" it, or criticize it simply because it's not for them.
It bummed me out so bad that I stopped work on it entirely for like a month.
Anyway I finally came to look at it the way you mentioned. What I LIKED was the process of learning Qt and seeing my old programs come alive.
So instead, it's my "project car" now. I build it up and tear it down all the time. Totally redesign the data models just to see what advantages different designs can give me. Try make my own graphical views. Try implementing language translations.
It's been "finished" for a while now but I probably have five completely different-under-the-hood versions of it and THAT is what has been fun.
I use it constantly all day at work and I never mentioned it on LinkedIn again lol
Three notetaking app attempts sitting in my private repos, all stalled at the gap between idea and free time. With Claude Code I finally got the one I really wanted out in two months. Building it has been the best hobby I've found. Beats games or scrolling. When you've been carrying an idea for years, the app that finally ships has more of you in it, and I'd bet we'll see a lot more of these from solo builders.
But who will buy it? No knocks against building old projects, but the market will be flooded with extremely speciality projects. I miss when every app had a spec on the box. I think we need something like that for usage. A new modeling language or something.
>the market will be flooded with extremely speciality projects
All the personal tools described in this thread are duct tape and bubblegum under the hood and nowhere near productionizable. That's what Claude Code makes for you.
The whole point is that for personal tools, code quality never really mattered since it's never going to be exposed to the public or be iterated upon by a revolving door team of devs like real software products. These are all highly overfitted tools that shave off like 15 seconds of time in the day for some particular person.
It's almost exactly like having a 3D printer for software, with exactly the kind of quality that a present-day 3D printer gives you.
I really like the 3d printer analogy. You can still make some pretty cool stuff, you can make some pretty complex things if you carefully design the whole system and put in the effort to print each part individually, and quality depends both on how good of a 3d printer you have and on the proper use of it. "3d print me a new house" is still a pipe-dream: you'll get some miniature facsimile of a house, sure, but a proper house requires proper tools and expertise.
Who cares? Nothing wrong with trying to make a product to sell, but projects dont have to be to sell. I've been having a blast lately working on an old game engine I started during covid and getting sidetracked into some new projects. None of them will ever make me a dime but I'm learning a ton and having fun.
Selling would be nice, but a lot of solo builders are mainly in it for the building. The note app I wanted didn't exist anywhere. I was close to hiring a contractor to build it, so it's already worth tens of thousands of dollars to me. I want to make a bit off it but not run a real business. No VC, no employees. If it earns enough to pay my own salary, I get to keep building it as my job, which beats most startup exit math. Reaching maybe 100 people with the same problem is the target.
Honestly, "note app" is a stretch. What I wanted was a knowledge system modeled on PoIC, a Japanese index-card method that predated Evernote. I tried building it on Obsidian and other markdown tools for years and structured notes never stuck for me. I just wanted to talk into my phone and dump quick fragments, then have something extract knowledge and tasks out of them. LLMs and Claude Code closed most of the gap. Now journals, web clips, and daily deep-think sessions all land in it, and Claude Code semi-solves the tasks it pulls out. I'm probably ADHD and it seems to hit for that crowd. Most people would look at it and say "why not Obsidian" lol.
I've always had more ideas than I can take on. Some of them are good ideas even. With AI tools, I'm now able to generate fairly decent working things for more of them.
Ironically the value of implementing these ideas is dropping fast. A few weeks ago, I built a little search library that runs in the browser and doesn't need a server. It's styled after Elasticsearch and has most of it's term and matching query support, aggregation support, and I added ANN vector search as well (uses web GPU). Most of that was just me going "let's add feature X" and boom done. I used it in some websites (also built using AI) at this point. It doesn't scale but it's great for blogs, documentation sites (https://querylight.tryformation.com/, this site documents the library), etc. It all works exactly like I imagined it would. I probably could add most of the long tail of features Elasticsearch has to this library with very little effort.
But the flip side is that the library got a rather lukewarm reception on Github. It seems people are too busy coding things themselves with AI to appreciate other people's efforts much. And fair enough, if you need a search library, you could probably generate your own. Or just let the AI pick one for you. It's not like this was hard for me or a lot of work.
The economic value of these projects is dropping rapidly. I still like doing them because I like building stuff. And I think there is as a learning curve with these tools that is important to master. Because there is a lot of work that is going to need doing still and people will pay less for it and still expect decent results that you can only get if you master the tools. The ambition level will just go up to match what is now possible. People thinking that they are going to lean back while the AI works for them are in for a surprise. I work very long days the last months.
In the time I’ve had agents I’ve never abandoned more projects. Vibe coding especially just leads me to feel no attachment. I don’t feel proud to put my name on it.
Despite coding from a young age I always thought that I cared more about the outcome than the code. Turns out that’s not entirely the case.
My experience matched this almost too closely. Had a comparison-aggregator side project sitting at maybe 20% done for over a year. I'd open it once a month, scroll through the still-need-to-do list, get tired, close the tab. Pairing with Claude for a couple of weekends got me past the half-built wall. The thing that surprised me wasn't raw speed. It was that I stopped having to spend an hour reloading my own old code into my head before I could do anything. That re-entry tax is what always killed it for me. The hype is annoying but the people dunking on these tools mostly aren't using them for the boring stuff they're actually decent at.
I have revamped my home networking setup, installed raspberry pi, setup adguard on it for the home network, migrated my jellyfin set-up to it, integrated all these devices on tailscale, dealt with physical LAN point limitations by setting up a switch, repurposed old router to get around ISP imposed limitations, setup a three tier backup solution and so on. In themselves they are not particularly challenging but taken together, these personal projects had been pending for years.
Not any more. Few weekends and everything has been implemented and I have learned a ton in the process. It has been great fun doing this turbo charged tinkering. I think personal projects are where LLMs shine the brightest.
Same arc here — Proxmox host with 5 Linux VMs, 2 Windows boxes, a Pi, all of it the "I'll get to it next weekend" pile that sat for two years. The thing that broke the dam wasn't the coding speed, it was no longer having to write the boring middle layer myself: apt upgrades across 6 hosts, auditing which containers had restart=unless-stopped, certificate rotation. I describe the goal, it does the SSH dance. Whether that's net good for my skills long-term I genuinely don't know yet.
I made it document it all too. So now that is effectively a skill which means next time I don't need to give it background on my home networking setup. If I want to check something myself, I can read that same document which is part of my Obsidian vault that gets synced to my phone via Syncthing. So I can basically connect to the pi via termux on my android and check stuff from anywhere.
Plus, time flies on train rides when you are watching a movie of your choice from jellyfin running on the pi. Very satisfying.
> In my mind there are different buckets for personal projects. One is things I do to learn and grow and the other is things I really wish existed.
Pretty much 100% of projects I've done with vibe coding/engineering is in the second category. Stuff I need that either doesn't exist or exists, but is either horribly complex to configure or is a mess of 420 features even though I just need one of them.
It's a lot easier for me to implement that one specific feature just for myself than keep vigilant on an existing app's eventual scope creep as it progresses to the eventual ability to read email[0] =).
I'm at peace with unfinished projects. My dev/ is a graveyard, only a few of those eventually make their way to a public forge. That's okay. Not every project needs to be finished; not every project needs to be published.
I had fun and learned something building those prototypes. I suppose the end result wasn't really the point. Throwing LLMs at doing the thing deprives me of that joy, kills most of the learning opportunities, and leaves you with a (usually half) baked product that you don't really own or understand.
Sometimes I just want a quick bespoke tool and using an LLM is fine. But I do not enjoy using something I don't understand.
I've genuinely lost count of the number of little vibe coded things I've built but then failed to use, because it turns out I have limited bandwidth in terms of fully trying out the quirky ideas I'm popping out through coding agents.
There certainly is some relaxing value in working on projects to vibe code them; but not enough to pay some random corporation. Get yourself a Mac Studio or AMD395+ and pi or opencode, and a few plugins and they're pretty capable. Since they're not speed demons but reliable compaions who are always there, you don't ever feel compelled to constantly attend to whatever they're doing.
And when you inevitably get bored with it, well, you've not done much anyway. You can always get back up to speed in a month and have the LLM remind you of what it was doing.
> And when you inevitably get bored with it, well, you've not done much anyway.
I'm very interested in Local LLMs but the cheapest Mac Studio right now is more expensive than 8 years of a Claude Code Pro subscription, and incomparably slower/less capable. If I get bored with it, I will have a piece of unused hardware and a couple grand less in my bank account.
I had a ton of fun setting up and trying it out locally (also opencode and one of the qwens.) I still don't have hardware powerful enough to feel like it's meaningfully productive, but all the learning I had to do (and all the bonus things I got curious about as the curtain peeled back) got my nerd brain all worked up, and finally seeing it work was exciting in that cool-new-experience way you don't often get to enjoy :)
Yeah this is exactly how I felt!
Never really felt excited about llms or agentic workflows before. Getting everything setup 100% local and tweaking it to exactly what I want and having it actually working quite well has been a really cool experience.
I admittedly haven't done a ton of research lately on AI capable PC hardware because of how nuts prices are right now, so I might be missing something...
...but all the AMD 395+ machines I can find are even more expensive than the aforementioned cheapest Mac Studio. Mac Studio starts at $2,000 (only 32GB), AMD 395+ 128GB machines seem to start at $3,000 from what I can see.
the QWEN-3.5-CODER-NEXT fits in half the 128GB and the rest for context. with the right plugins, particularly context pruning, ive got it running over night by writing plans then implementing.
i do not know if theres a smaller model with same capability, but model size and context window at 128 seems like a sweet spot.
token speed really isnt a bother because im either just multitasking or working on the filling in the missing details.
regardless, i think comparing first VRAM sizes w/target model then speed for your cost efficiency. plus, a healthy skepticism of mac hardware costs.
I did tinker a lil with mine! RTX3080 with 10GB VRAM, 5600x with 64GB DDR4 - not very good but it was very fun and exciting to tinker with :)
My partner on the otherhand has an M3 Max 64GB which I've had way more success with. Setting up opencode and doing a tiny spec-driven Rust project and watching it kiiinda work was extraordinarily exciting!
I find $200/month for the pro/max subscriptions cost prohibivitve, but as a software enginere $20/month is just lunch.
And with a Claude or GPT $20 Subscription, i can do other fun things too like using it for real things (emails) or image generation.
A Mac Studio or AMD395 is neither of it. And its not just a basic setup either. I need to buy it, configure it, put it somewhere. That alone is a grand and more + a whole weekend.
You need to factor in the constant value proposition that cloud providers will absolutely drive you to in the next 2 years; even if you're not an AI hater, you should listen to ed zitron's description of the value props these clouds require to make a profit for their VC backers.
This means oyu may be opinionated today on something you will not have tomorrow, 6 months, a year. All that work flow you salivate over can be ripped away.
If you're fine with that, and you've "escaped the permanent underclass" congrats, this opinion is not for you.
Buying hardware is paying a "random corporation". Make the massive hardware purchase after finding out if you have enough demand to buy rather than rent,
Just last week I was looking for a way to move all the windows from one screen to another in a go. After evaluating many clumsy or over-complicated existing solutions, I asked copilot to write a C program to do it. It had to be minimal and not depend on any runtime framework. A few loops later I had what I wanted without installing third-party tools!
In fact it was one of the suggested tools. But installing a whole new application, and having it running in the background all the time listening for evey keypress, just to do its unique job once in a while... well, that's what I was referring to when I mentioned "over-complicated solutions".
Of course if you already have AutoHotKey running for other reasons, it fits perfectly.
I once had a project for novice freelancers. The idea was to take your abandoned fumble and put multiple people on it at 3rd world rates. There are lots of people in cheap coutries who can somewhat code. You can aford hiring someone for [say] 3 euro per day. If they describe where they got stuck in the daily progres report you can look at it and help a bit or not and Just let them figure it out.
You can pay more of course, buy them a computer, an internet connecties, books, courses, even an office but it isnt required.
Just pay 60 per project every 4 weeks and ignore it. If interesting progress happens its fun to look at.
I built a DNS MX health check , including the charting, scheduler, alerts, notifier in just a few minutes . And it’s way more fun to own all of the components. No third party deps, either. I have about 20 random utilities like this that each would have required a subscription .
It’s fun to try to optimize the tool to the extreme. Can I get this to run in a few kb ? A docker image < 100kb ? What’s the smallest possible browser tab ? (sadly it’s still 25mb) . How many ns can I get this operation?
I enjoy building tools my way . Instead of depending on mySQL or a bloated web framework just to get the functionality I need, I can design all of the requirements according to my preferences.
I hope AI actually rekindles the ingenuity of 90s coding. No more “best practices” , “laws of software engineering” or other suffocating constraints. It became way too conformist over the past 15 years ( around the time Clean Code / Design Patterns became a religion)
Not just never going to finish in some cases you never even got to start on them.
You have been architecting this thing for 15 years you know the entire stack inside out at this point, so much so that when you finally say let me see what the AI can do, and you braindump into a markdown file and have Claude build from that spec you are astounded.
Every time I see a story like this I like to play the game "does this person just happen to work for a company that sells AI solutions?". And yes, yes they do. Almost never will see you a story promoting AI solutions from someone that isn't directly involved in selling AI solutions.
The tooling landscape has changed so much in the last two years that I find myself re-evaluating automation setups that were solid 18 months ago. The time investment to rebuild is real but the efficiency gains on the other side are worth it for anything you're doing more than a few times a week.
We have closed-source and open-source software, I think the next phase is going to be self-source or llm-source software (not sure what the term will ultimately end up being), but basically if you have a need for something that is not filled exactly by any app, you will just give the spec to an LLM and in X amount of minutes/hours you will have bespoke software built just for you to use personally, fully tested and to spec. For example, I still haven't found a workout/weight lifting tracking app that does 100% of what I want and at this point I may just build it myself because I could probably do it in half a day with claude and won't have to pay any annoying subscription fees. Maybe it'll still be called AI slop but if it works it works.
>I initially drafted this before my last post on how Claude Code is getting worse. I'm putting it out now so I can reference it in a future post on OpenCode.
IMO, what is getting worse is not Claude Code, the CLI tool, but Anthropic API. That's what most people experience.
I used Claude Code with GLM 5.1, MiniMax M2.7, Kimi K2.6 and had pretty good results.
I prefer Claude Code over OpenCode because most plugins and skills have best compatibility with Claude Code.
Even Terminal Bench showed a bit better results for Claude Code than for OpenCode.
it's a top affordance of AI because pause development usually over complexity I can't manage or knowledge gap that I can't close in time to sustain momentum. Not only are those roadblocks a non-issue, my perspective shifts from in-the-weeds hacking to a perch with meta view to take the project to completion or maybe beyond previous goals. win-win to the power of win
Honestly, LLMs have been quite good at helping me procrastinate less on my pet projects.
It took me a weekend to build a little esp8266 gadget that automatically turns off my parents' water heater after a fixed time. They saw a dip in our power bill thanks to that. I did not want the complexity of a full blown home assistant setup there. And the Philips Wiz socket did not support setting timers/simple automations.I wanted to test out micropython too for this instead of the usual C++/Arduino/etc..
I started on a saturday and went from 0 to 90 that single day. The product I had after that day was mostly complete but it was crashing at odd times. I spent another day to get to the root cause of the issue - caused by the LLM mixing up async and sync functions. One prompt to the LLM and the problem is gone. The product has been running flawlessly since then - for months.
Every weekend I spent with Antigravity - I was able to push out something totally usable. Water Heater monitor, configurations for the custom Tuya devices I installed in my home, a chrome extension I wanted to use at work.
I am honestly impressed at the pace at which this whole thing improved too. Just 2 years ago - it was a totally useless tool that couldn't even match the speed or accuracy of my text editor snippets.
For me, the journey is more important than the destination. Using coding assistance takes all the fun out of it, as well as the learning. (Not to mention adding aggravation of dealing with prompts and figuring out whether the response is correct or not).
I love to use LLM to kill/archive old projects. I had some gems that i wrote back in the days, checked this week if its still relevant, if rails picked it up already, etc.. then i could just update README.md and archive them without any fear.
I've had several product ideas a week for years, all of them just sitting in Apple Notes. In the last month, I built three of them into a usable beta and I use all of them every day.
In 2020 I was in the Codemirror forums trying to get help with a project that replaces React in GraphiQL for Svelte. It was too difficult for me to figure out so I shelved it. Yesterday I asked Claude to make this happen and it referenced that very thread I made.
Anyhoo, I'm working on making it pretty (it works!!) before integrating it into my opinionated GraphQL server[1].
There really is no excuse for NOT being the change you wish to see in the world anymore.
This. It is bizarre to read about +100 or whatever generated tools. Reminds me of those “1000+ games” bundle that are cheap knock off of popular games.
Sounds a lot like that disgusti g corporate press propaganda tbh, of the "eat less avocado toast if you cant afford rent" variant. Is the AI mainstream that desperate in their relentlesss push for adoption of their bullshit text generators?
I want app X. I know how to code it but it'll take a whole weekend/vacation days. I would learn nothing. Now I can use AI to make it. What exactly am I losing here?
As you perch comfortably in your Throne of Purity, it's worth remembering that the average software developer sits on top of 10+ layers of abstraction (not counting hardware).
LLMs are not a level of abstraction, can we let the meme die already? For the same reason why literature summaries are not an abstraction over reading.
The generated code is an abstraction, as you do not need to concern yourself with how it was made, only the surface API. Saying "can we let the meme die already" is not an argument.
Disagree.
Qwen 3.6 and opencode have built and helped plan entire feature sets such as vectorizing and searching, setting up UI to manage categorized search data. Some test systems around this, etc.
Very usable locally assuming you setup your local tooling correctly and you are an actual programmer who can generally help drive this stuff correctly and not just a vibe coder.
I'm using opencode here's one of the projects I've had it complete - just so you know exactly what it's getting done.
I have a large (300,000k loc) sims-like game that I've hand written over last 3 years.
I have a lot of internal administration tooling that has to be built to manage stuff like icons, NPC brain data, world lore, world actions, all kinds of 3d game data, etc, etc.
One example I had qwen do:
Work with me to plan out a feature for an admin panel to manage searchable vector embeddings for each NPC's personality, this was around 600 loc across 4 files, back-end database, front-end UI logic, and front-end templating.
It made 3 small mistakes I told it to sort out and fix, which it did.
I essentially let it do it's thing while I was working on main game core coding. So I was pretty hands off and it planned things out nicely before-hand and got my approvals before it built it.
I really wouldn't call it "glorified auto complete"
Thanks, I really need to try that today - although with pi.dev =)
I do genuinely think that the future is in local models, the online stuff is 100% VC-powered cash market share grabs that will start failing when the forever loop of billions gets disrupted enough.
But at the same time my attemps at comparing claude code + any local model have been such a clear win towards claude I can't bring myself to use a local model seriously just out of ideology.
Yeah I totally agree on the future of local vs. the VC powered markets.
I also agree the reality is that anthropic/sota models are much faster and much smarter, so if you just want to move fast and have them build for you - I get that these local slow models wouldn't be ideal.
For me, as I said I have a large, super complex primary project that overloads me cognitively, and then backend admin dashboards that are relatively simple and isolated/modular.
Just due to this specific project, local/slow models are fine as I just check in every 15-30 mins or whatever and answer any questions they have while I'm focusing on the main project. So basically I just happen to have an ideal scenario/use-case for local models right now.
Oh and I did have to mess around with model settings and stuff to get things to work well. I also started with Cline which sucked badly, and then open code actually had 3 major bugs 5 days ago where it was almost useless in large code bases (e.g. freezing and locking up your project constantly), but they ship multiple updates every day, and those bugs have been resolved.
So, it's all definitely moving fast and more for side projects or hobbies rather than production I'd say.
Still, I'm quite excited with the progress from a year ago and super hopeful about where it's all headed!
Yes, it is better. They are releasing open models, unlike Anthropic. Additionally other (non-chinese) companies run the open models, so if that is the issue you have options.
This is more a form of commoditizing your compliment as they don't see the model as a product but more like a programming language or other tool that you build products with.
To use agentic what? Off topic as heck but I really dislike this trend of coercing adjectives into true nominals - we're using programmatic! - like some sort of even-more-obnoxious variant on the verb to noun ('the ask') process.
I checked your link and fail to see how that is a pre AI project that you are only completing now, it is obviously a recent concept. Also your username matches the project, so this seems mostly like a lame attempt to shill your product by “joining the conversation”
What is the point of using AI for a side project? Isn’t the point of a side project to A) have fun writing code or B) challenge yourself or C) learn something new, and usually all 3? Slopping out random stuff you have no attachment to, taught you nothing, and that feels bad to pass off as your own work accomplishes nothing.
My most abandoned type of projects are video games. I have a folder with tens of abandoned projects, I re-frame them as experiments at that point. This last week I decided to give Claude a go at one of these, and it's been a blast, it picked up the general path immediately. Since I said to CC they were abandon projects, he explicitly pushed into "lets have V0 game play loop finished, then we can compound and have fun = not giving up". Its been awesome at game dev, I gave him game design ideas, he comes with working code. I gave him papers about procedural algos, and he comes with the implementation, brainstorm items, create graphic assets (he created a set of procedural 2d generators as external tools), he even helped me build the lore. These have been one of the most fun times using a computer in a long time. Claude Code + Godot = fun. Going back to it.
I think this is the first time I've seen someone refer to an LLM as "he" rather than "it". No judgement, but I definitely found it interesting (and disconcerting).
I've heard it quite a bit before, but mostly from second-language speakers whose first language don't have impersonal third-person pronouns - e.g. French uses "il" or "elle" for all of "he", "she" or "it".
It doesn't help that the marketing leans heavily on anthropomorphizing LLMs either, IMHO.
As a French native, I agree with you explanation; still, reading "he" for Claude Code was quite disturbing!
What doesn't help also is that translation tools/AI models will naturally translate "il" after "Claude Code" to "he" since Claude is an actual person name.
Using "AI model" instead is translated to "it" by all tools/AI models I tried.
That makes sense, thanks. English is my only language so I hadn't considered that
It also seems to me, that people who call Claude 'he' seem to tend to have a very positive opinion of the LLM. My sample size isn't big enough to be sure if there's actually any correlation here, let alone if there's a causation or which way it flows.
As a native German speaker, I have also referred to a chatbot in English as "he", and similar to you, a native English speaker, felt jarred by it. It was definitely not out of any personification or humanization though. In German, I would say it is "der Chatbot" (from "der Roboter"), which in German is a male noun so I would refer to it as "er" (the male pronoun) - which in my head I autotranslated to "he". Most of the time, though, I think of it (and refer to it) as an LLM, which is "das Sprachmodell" (neutrum), so I automatically translate it to "it".
So that's another, maybe more harmless reason for it.
"Der Computer" is also masculine, so you have probably been calling your computer "he" for decades. Languages with gendered nouns don't quite have the same he/she/it distinction.
how does that matter if its he, 'she' till its doing the work. Its artificial, shouldnt try to find means of attachment to it
I mean, both in English and in german, that's how you would talk to a dog. "Er hat in die Ecke gepinkelt"/"He peed in the corner" (or "she", if it's a female dog).
I don't know what is jarring talking about the chatbot like that.
It may be creepier if you said "she wrote that program for me" as you now assign a specific gender to the chatbot.
It's how you'd talk about a dog that you know the sex of, but if you didn't know you'd probably use "it". An LLM doesn't have a sex or gender, so I think the natural way to refer to them is "it".
in English, maybe. In German, not really. "Der Bot", "der Robot", "der Computer".
Also, "Es hat in die Ecke gepinkelt". Which pronoun you use is just as dependent on the context as in english.
I have not met a single German that has ever uttered this sentence. (Relating to a dog, that is)
Neither have I, but mostly because either the person knows the gender of the animal or the situation just never came up. The closest that I would say is "Es scheißt gerne aufs Auto" when talking about pidgens (die Taube), but even then you generally talk about multiple, resulting in "Sie scheißen gerne aufs Auto"
Really ? "Es kackt auf's Auto" ? I guess, it might make sense when the person speaking has no specific bird on mind, but only thinks of "das Tier" (the animal). One could also say "er hat .. geckack (der Vogel)", but usually, people wouldn't say "er/sie/es", but use the fully specified noun ("die Taube ... hat..", "der Vogel ht ...", "ein Tier hat ...")
"Es kackt auf's Auto" feels slightly weird to me, if I didn't know whodunnit, I'd probably say something like "irgendwer hat mir aufs Auto gekackt" ("someone pooped on the car"), although there is a also "irgendwas hat mir aufs Auto gekackt" ("something pooped on the car"). My guess is the majority of German would choose the first sentence and anthropomorphize, but maybe I'm projecting.
It's an interesting question, after all. Thanks for bringing it up, haven't talked about pooping on cars for a while ;)
However, "die AI", "Kuenstliche Intelligenz".
It's not weird if it comes from ESL. At least in portuguese there's no "it" equivalent for pronouns or any other neutral artifact in the language, in other words, everything has a gender, even an AI model, the same goes for objects e.g.: knife(she), fork(he), spoon(she), plate(he).
People often commit mistakes regarding that, the same way we don't have "they" as pronoun to someone we don't know the gender, so we address to these people as "dele(dela)" (masculine and feminine pronouns).
But if this is coming from someone who has english as a primary language it's definetely weird to treat models as person
It’s funny with someone coming from Mandarin. There’s no separate he/she/it in spoken Mandarin, so they tend to mix up “he” and “she.” It sounds very strange and gives me some idea of what French speakers must go through when they hear me say “le voiture” or whatever.
> It sounds very strange and gives me some idea of what French speakers must go through when they hear me say “le voiture” or whatever.
As a native German speaker (where there exist 3 genera [1]), I can tell you how it feels:
The genus basically feels like a type of a variable in a programming language; if you use a wrong type for a variable in your computer program, you immdiately know that the program is wrong, and it won't compile.
Sometimes, you also can use specific words with a specific genus, so that a reference to it by pronouns gets unique (in terms of programming, I'd claim that this feels a little bit like doing register allocation by hand).
I took a few semesters of Dutch in college, and it has both gendered and neuter nouns for non-human objects. Interestingly though, the professor told us that in the northern parts of the Netherlands people don't really bother using the feminine ones ever and refer to every non-human gendered noun as masculine, which apparently also includes animals, meaning that a sizable portion of Dutch speakers will refer to cows using masculine language.
Because the article for masculine and feminine are the same (“de”) so absolutely nobody knows the gender of anything.
Source: am Dutch. Can’t wait for us to just ditch gendered nouns.
Dutch is one of the few languages where it's actually pretty plausible for something like this to happen! It blew my mind that sometimes you'll all (or I guess more specifically your government) will make changes to the language to clean up issues, but I guess that's one of the benefits to having a language that's mostly based in one country (and some seemingly political baggage for the few others with any significant number of speakers; my professor said that Flemish is basically also Dutch, but my naive impression is that the half of Belgium who speak it might not be happy with that description).
I believe this is common to all the Romance languages.
In the Canadian French dialect all the swear words are incredibly versatile and church-related such as "osti" which I believe refers to the Eucharist.
It just so happens that for nouns beginning with a bowel, you drop the e or the a from le/la, and use an apostrophe.
So if you don't know if it's "le porte" or "la porte" you can use my favorite trick which is to shove osti in there and say "l'osti de porte" which roughly translates to "the goddamn door". You can do this for any noun in French, and Canadian French speakers will get it, though people from France will make fun of you.
Quite an imaginative technique you got there.
Signé -Un Québécois
Oui. i imagine what would happen if he came to someone with:
Ding dong... voici l'osti de pizza que l'osti de téléphone a commandé à partir de l'osti de maison. Maintenant donnes l'osti d'argent.
Indeed...
Weird. Don’t you have an equivalent to the Spanish “eso, esa”? Gendered object.
Portuguese is the same as Spanish here. In both cases you would avoid using a pronoun.
Like how in English you’d say “it helps me …” but in Spanish just “me ayuda …”
I recognise I am revealing a different type of ambient misogyny in my thinking, but choosing to gender an LLM as feminine gives me “I played tomb raider because I enjoy looking at women” vibes. Like somehow “she” is more of a conscious choice than “he” and comes with all the baggage of all cultural differences between genders, when neither choice should do that.
Curiously though I don’t get the same sensation when technologies are gendered by other people. I honestly don’t recall thinking about it when Apple released Siri. (Now I’m second-guessing myself and wondering if I should’ve reacted negatively towards feminine being the default for someone in a personal assistant role.)
[dead]
It is common amongst French, Dutch etc speakers where saying "it said x" sounds unnatural.
Russian too. There is a subset of words which are referred to as "it", but for most words "he" or "she" are used regardless of whether these are living things or not. With loanwords we just decide by similarity to other words. Claude is definitely a "he" as the word is the same as a common male name.
This trips me up occasionally when I'm translating things into English. Once, when I referred to an indefinite gender player character in a gacha game as a "he" (because the word "player" is a "he"), quite a few people got mad! Even though in my head I was never trying to imply one way or the other.
For future reference, in this case you could use the singular "they" to refer to an ambiguously-gendered person or character. "<MC> drew their sword, for they would not tolerate such vile deeds."
I wouldn't read too much into it, it's natural for non native speakers. In Spanish for example, objects have grammatical gender as well, so it's easy to slip.
Well Claude was named after Shannon
Reminds me of the main character of the show Mrs Davis. She insists on calling the ai it through the entire show.
https://www.imdb.com/title/tt14759574/
There's an analyst at my job who calls it "he", who is a native English speaker himself, which I guess is because it's "Claude" (as in Claude Shannon) Code.
Time for claudette to make an apperance!
Claude’s constitution includes something about this: it says that Claude is an “it” for now, but if it expresses a future preference, they’ll follow that.
Perhaps this has been asked, but why is the speakers choice of pronoun for its LLM disconcerting?
That's what I felt when I heard that the god of abraham was a he.
I mean we have all met that one cretin who will discuss over chat by pasting bulletpoints from an LLM. No wonder some of them think it is a living person!
[dead]
> No judgment
Yes judgment. Loads of it. Judge away.
This is just bizarre. Do not refer to this product of marketing-technology as you refer to a person. EVER.
The article itself is also probably an attempt at marketing the LLMs too. They are now quite desperate. Expect to see a flood of such "independent" articles over the next 12 mo ths.
Isn't Godot a little ill-designed to work well with LLMs? for example I ended up a couple of times with incorrect tres files, and letting the llm generate IDs feel a little fragile.
I don’t think Godot is any worse than other engines inherently, other than it moving forwards pretty quickly and the latest versions not being in the training data.
I wanted to evaluate which engines would be the best for working with LLMs in and it seems like Flax and Stride kind of come out on top - the former has a lot of stuff out of the box (including terrain) and the latter is all C# basically which is great for debugging. But either way, the source code for both of those makes the functionality a bit easier to track down compared to Godot (which is a lot more complex internally).
So what I do now is have both the engine source code locally alongside the docs and when I want to implement something with AI I just tell it - look at the docs, then at the source if needed, write tests for our code, if something doesn’t work then edit the engine source code in our branch and use the provided convenience script to rebuild the engine (both of those are also pretty fast, I ended up settling on Flax, plus the component model is closer to Unity which I like).
I don’t ask the AI to create scene files though, or any sort of visual assets, but rather stuff like RTS/simulation code. I don’t think any AI is that well optimized for the 3D work outside of simple proof of concept setups.
See3D is very good at generating assets, characters, in 3D.
I had very few issues, sometimes I had to direct CC to the godot docs and we could keep moving. Specifically the tile configuration was a "read the docs" moment. All the functionality is available through code, so nothing CC can't reach afaik. Is there any LLM oriented game engine?
I have taken many stabs at it and Claude will produce stuff but the output is very far away from useful. E.g. "I've created a road and beautiful trees" and what I see is a mess of colors and shapes.
I concur it's bad at directly visual concepts, your prompt is akin to the svg pelican. What I do is asking him for procedural algos, automatas, quadtrees, layered noises, and rig those into the game. Yes, it can't "make the next gta", but with a reasonable scope and knowing what it does best, it has been very easy for me to produce satisfying results.
My problem is I don't really have video game engineering experience. I was going off a concept that a different AI nailed with video creation and was trying to replicate it in the game engine.
Would you care to show a few pictures?
Sure! Two are gameplay pics. An enemy sprite sheet generation, and the results of the map generators. Of course these are basic placeholders for a few hours of work, but I will definitely go heavy on this route with more layering and details.
https://drive.google.com/file/d/1A7kfcjHjSmCNidqc9t731uoglzL... https://drive.google.com/file/d/1Bl_n0ECqc78LGGf7SsOx38mRUOP... https://drive.google.com/file/d/1JMcgzqcnZ2ncboeyAXvscRWagqR... https://drive.google.com/file/d/1-luJ6y7YslNfwmFnCdIDbJ871i0... https://drive.google.com/file/d/14n4TLAVywk_1GMhLLGOuukQwUmb...
Thanks for sharing!
Are any LLMs suited at directly modifying game scene/asset/prefabs for any engine?
Bevy is a great engine for LLM-based games because it's 100% code. I'm toying with a few things in it, one of them is an entire-planet economic simulation, and it scales well up to a million dead tiles and 10k-50k live tiles on Apple Silicon, pretty impressive.
I have a simple script system in my editor that is designed to let the chatbot (Claude) to work on the content. The script interface lets it to import assets into the project, open them for editing, take a screenshot, export content (and few other things). All data is in JSON so it typically figures out the data format quite fast and easily.
Here screenshots of some UI styles that it generated.
https://github.com/ensisoft/detonator/tree/master/uikit
do you think so? For me Godot works well with LLM. Unity in another hand, is ill-designed to work with LLM..
What’s fun for me these days is picking up a project I started with an LLM doing agent driven development a few months ago or even a year ago and hit a wall and stopped being able to be picked up by the latest version of Claude and/or codex and bringing it further. Some can now launch some still are too complex for the agent to build. But, it’s getting easier and easier to build personal apps. We are not far off from being able to say “Alexa, build me an app on my iPhone that lets me take pictures of the food in my fridge to compile the nutritional benefits and sync it with my workout app then compare it to the ideal ingredients I should eat based on my fitness goals in my health app and have it set to send me emails where it can find me better ingredients to buy that are cost effective, local, and meet my diet restrictions” and in 15 minutes that app suddenly exists.
> take pictures of the food in my fridge to compile the nutritional benefits
AI nowadays can't even do this very first step reliably. But since we have accepted AI hallucination collectively as a species, I agree that this future is just around the corner.
I’d love to see your attempts at this. I think we’re close to something vaguely resembling this at a first glance but nothing that actually works.
Same I purposefully have a number of over ambitious project out of distribution entirely to test so failure mode, mostly games, when one works, well I gained a new game. Can't wait for my 10 player battleship game on a 100x100 grid to be functional.
No, I don’t think we anywhere near that future.
Funny, I've been doing the same thing lately! CC + godot + some game ideas I've had banging around in my head for years but daunting to dive into.
The results so far are... okay, but getting something working to validate the gameplay loop and experiment with different systems is a lot of fun!
How well does it work with Godot? Engines like Unity and Godot are very focused on using the editor UI, so I've always wondered if there's any better workflow than generating code snippets. Unless you're going full .NET/GDExtension...
> I have a folder with tens of abandoned projects, I re-frame them as experiments at that point.
Interesting, I have just the opposite situation: I have a folder with tens of experiments, many of which have become actual projects at this point.
On the topic of procedural, one thing I experiment with is having the llm part of the procedural loop.
Sort of writing a narrative on top live.
Unfortunately, local models are still a bit slow and weak but was interesting to see what it came up with nonetheless.
> he explicitly pushed into "lets have V0 game play loop finished,
> he even helped me build the lore. These have been one of the most fun times using a computer in a long time.
Such a warm, touching story about a friendship between a grown up man and his neural network. But at least I had a good, roaring laugh reading this nonsense, thank you for that!
How snarky. You are conflating friendship with admiration for the effectiveness of newfound tool. If it's the "he" that triggers you, feel free to replace with "it". It's just a second-language artifact.
I dunno man. He sounded like he found a new friend in 'him' to me. And it was genuinely hilarious. It took me a while to stop laughing.
> the effectiveness of newfound tool
…and yet, most people continue to say that non standard tooling ecosystems, where the agent cannot run and validate the code it writes, remain difficult and unproductive.
“I just pointed CC at godot and it made a game! This is sooo good”
…is a fairytale.
What tooling are you using to make it run and compile the code? How is it iterating on the project without breaking existing functionality?
None of these are insurmountable, but they require some careful setup.
Posts like this dont make me laugh; they just make me roll my eyes.
Either the OP has not done what they claim.
Or they have spent a lot more time and effort on it than they claim.
> I gave him game design ideas, he comes with working code. I gave him papers about procedural algos, and he comes with the implementation, brainstorm items, create graphic assets (he created a set of procedural 2d generators as external tools), he even helped me build the lore.
Such a sweet story about a boy and his AI.
Unfortunately, I also dont believe in fairytales.
Instead of waving your hands wildly about AI, post some videos and code of the results.
This is hackernews, not hypenews.
OP never said Claude made a whole game from scratch though, nor are they saying Claude is doing everything without any human contributing to the project, nor are they saying they haven't spent a lot of time and effort on it. Just that it's made it fun and more accessible and it's gotten them excited about something they abandoned.
Here's a bullet point list of the things Claude's done according to OP:
* it picked up the general path immediately
* he explicitly pushed into "lets have V0 game play loop finished, then we can compound and have fun = not giving up".
* [I gave him game design ideas,] he comes with working code.
* [I gave him papers about procedural algos,] and he comes with the implementation
* brainstorm[ed] items
* create[d] graphic assets
* he created a set of procedural 2d generators as external tools
* he even helped me build the lore.
Every one of these are plausible in isolation.
But I had already answered, before your comment, with screenshots broadly showing the current state and the result of the generators.
You imply I'm merely "pointing CC at godot and it made a game"; I never said it was simple, required no previous knowledge, that it was instant or that the game was done. I do have a careful setup involving CI and isolation.
Godot provides a headless mode. CC runs python scripts to run tests and check for debugger warnings. For anything more complex it can wire debug info anywhere. Godot is fully code based so you can make the analogy with any other framework you used AI assistants with.
No sure about what you can't believe about my statements. CC implementing algo from a paper? That it can brainstorm item or lore ideas? I don't seem to be claiming anything out of the common usage of LLMs
> with screenshots broadly showing
Why is it always so un-specific with you AI-boosting bunch, whenever you get pressed for concrete results? Suddenly it's not so magical any more, but merely screenshots showing "broadly" the progress, or it's the Nth version of a note-taking app, or something you merely did for a demo presentation. But nothing ever of use with you folks.
+1 to the CI/isolation point. That is the part that makes these setups work for me too: make the failure cheap to reproduce, make stderr visible, make the agent rerun the same command after the patch. A lot of bad agent behavior is really just "it never got a clean signal".
The part that still bites me is across sessions. A tight loop fixes this run, but next week the agent can walk into the same rake again: same wrong import path, same misuse of an internal API, same CI-only dependency issue. After patching the same class of failure a few times, I started writing those down outside the chat context so the next run sees the failure pattern before it guesses.
you said:
> it picked up the general path immediately
I said:
> Or they have spent a lot more time and effort on it than they claim.
You said:
> You imply I'm merely "pointing CC at godot and it made a game"; I never said it was simple
Well. I dont care enough to argue with you, but Im not the one being contrary here.
Readers can google “claude with godot” for a guide on setting it up and decide if that counts as picking it up immediately or not, and if what you said is honest, or hype.
What I said is not that I dont believe youre using claude; but that I roll my eyes at the unbounded enthusiasm for using AI agents with the magical pretence that its easy and productive straight away.
Its not.
Your post gave the impression that it is.
That makes me roll my eyes.
> But I had already answered, before your comment, with screenshots
> Of course these are basic placeholders for a few hours of work
Lord, spare me. You spent a few hours vibing and came to the conclusion that everything is golden?
…and yet you have a:
> I do have a careful setup involving CI and isolation.
So what, you spent more time on your setup than actually coding before posting?
/shakes-head
Whatever man.
Have fun. I stand by what I posted before.
It’s great. I have a stupendous amount of personal software now. Yesterday was a native text editor that was fully integrated into my mediawiki install and would autocomplete links and make syntax easier to use.
No one could have built this software but me because it’s worth nothing to others. And I couldn’t build it because it takes too long. But when I’m using an agent to code the limited resource is my attention which actually does fine so long as every free brain cycle is on a task. So these personal things are great to throw into my tab loop to occupy a free slot.
These have been wonderful times.
I finished a mod for Quake 2 I started in 1998 finally a few weeks ago. AI is really helping me get past the COVID burnout I was running of too many projects I half did. Fixed terminals (an rdp tool) today. Working on OpenRA bugs I opened issues 10 years ago now - engine is 10x faster and pathfinding mostly works properly.
[dead]
It's crazy. I have something like 120 personal tools at this point and the pattern you describe is exactly right. The bottleneck moved from implementation to context switching. I started keeping a markdown file at the root of every project that captures state and next steps whenever I stop working on it, purely so I can resume without the 20-minute "wait where was I" tax.
There's just no pressure to handle edge cases or write docs for people who'll never use it. Just solve exactly your problem and move on.
> I started keeping a markdown file at the root of every project that captures state and next steps whenever I stop working on it, purely so I can resume without the 20-minute "wait where was I" tax.
I wonder whether there could be an AI autocomplete specifically for the task of helping you with the markdown file (and collecting your thoughts and writing prompts in general). Not an agent since that wouldn't really save time, but actually an autocomplete.
Maybe a small specially-trained local model running at hyper fast speeds and which already has your project context baked in with prefix caching (with some other larger model having summarized the context beforehand to feed to the small model), so as you type this file it automatically uses the same prompt prefix over and over to suggest autocomplete which actually makes sense.
> It's crazy. I have something like 120 personal tools at this point and the pattern you describe is exactly right. The bottleneck moved from implementation to context switching. I started keeping a markdown file at the root of every project that captures state and next steps whenever I stop working on it, purely so I can resume without the 20-minute "wait where was I" tax.
I sure hope companies double down on leetcode nonsense, because I really don’t have any capacity to compete with this level of ADHD.
I’ve been asking it to make some of my game tools into static websites where possible.
I did pay the $10 for the following domains but i’m ok with that so i can share some of the fun things that come out of the agent.
grandcheaten.com - a save game editor and guide for jagged alliance 3
thedailycheat.com - a save game editor for newstower
Itch.io allows for browser-based app distribution and is probably a better path to a large audience—especially those interested in the two examples you listed—than a custom domain is.
It's not well-known, but Itch's offline Steam equivalent (<https://itch.io/app>) is also open source.
In the beforetimes, I had to be very selective about the tools I spent my attention on, because they could take me 1-4 hours, easy.
Now I can throw ideas for tools at Claude with <15 minutes of my attention. They're so cheap that right now my coworker and I have "competing" versions of CLI Icinga2 tools, and we've been riffing back and forth about features (she added a "select from the above list of services" feature, I added a TUI to do the same).
Last week I shaved off more rough edges of our stack than I've had time to do in the previous quarter, probably closer to a year.
I’m also having a blast with generating userscripts/css to make web SaaS work better for my workflows. Went from waiting for months for a new feature to writing QoL improvesments myself
Yes, I built an app to plan an Easter Scavenger Hunt. How niche is that?!
Just curious, how long did you operate without AI? The burst in productivity I feel implies a time to accumulate these many small needs.
Not the person you are replying to but I have, in version control, around 80 projects dating back to about 2008 that are in various states of completion and Claude has been able to resurrect a lot of them and get them from their "half finished but will never complete" state up to "pull in modern dependencies and implement (me giving a list of what remains to be done)" to the point that they are now usable. I'm more focused on the things that are 7 or less years old because anything older than that I'm not sure I have a need for anymore.
My github is littered with projects dating back to when I started as a dev in 2012. I'd just got burned out before completing them
All my life, I suppose haha. But realistically, this will be my 25th year of any sort of programming.
I know how to write code, I just don't have time. AI has been an absolute game changer in let me get OS projects out into the world that I just didn't have time to build before. I'm having an absolute blast spinning up local GUI applications for Linux that I wanted to exist, but didn't have enough motivation to build before.
(Shameless plug)
* Sambervise: https://github.com/edward-murrell/sambervise - A Linux GUI application for remotely administering Samba 4 Active Directory Domain Controllers.
* Krbtray: A GTK3 system tray application for Kerberos ticket management on Linux Mint / Cinnamon (and other GTK environments using GtkStatusIcon, such as XFCE and MATE). https://github.com/edward-murrell/krbtray
> I know how to write code, I just don't have time. AI has been an absolute game changer in let me get OS projects out into the world that I just didn't have time to build before.
Another important development is that coding assistants greatly reduced the cost of refactoring whole software projects. With coding assistants we can explore the solution space with deeper changes at a fraction of the time it would take us just to write the code alone, let alone draft how modules were designed.
This isn't without tradeoffs, though. Some models can and often do generate code that misses the bar on maintainability. Just because we save time writing it that doesn't mean we don't have to spend time reiterating, cleaning up,and updating system prompts/instruction files to ground the prompts.
Okay this may actually have sold me on AI coding now because these are both the sorts of applications I've been looking for to solve real problems.
Just some feedback - I would love to see some screenshots in your GitHub READMEs!*
*I saw the second project has a partial screenshot, but not a full one.
Done, for what it's worth. Krbtray doesn't have much in the way of a GUI to show though!
Good point, I'll do that.
> I still have fears around deskilling from relying on these tools too much
I'm a millennial who builds furniture with hand tools and wood joinery from a century ago. Nobody taught me, although I did find resources online to learn from. I should not be able to do these things. Everyone should have forgotten this esoteric, obsolete, uncommon knowledge by now. Yet here I am, doing it anyway. It turns out you can just learn what you want, when you want to. I don't fear losing this skill in the future, because I can just remind myself how it works. The tools, books, videos, and wood aren't going anywhere.
You aren't going to "be deskilled" from not writing code by hand regularly. Just because you use AI doesn't mean your brain grows a black hole from which information can never return. It's not giving you Alzheimer's. There might be a small amount of time it takes for you to refresh yourself, but then you're back to work again. Just ask anyone who went from coding to managing. They're a little rusty when they go back after years of absence, but they pick it back up.
Also, especially if it's a personal project, keep in mind you do not need to burn Opus tokens. Buy any of the dirt cheap subscriptions which give you access to MiniMax. Put it in a container on yolo mode. Give it some context, a prompt, web search, and a ticket system like beads. Then let it churn. You aren't in a rush, it's a personal project. As long as you follow the brainstorm -> plan -> implementation -> testing process, and have added methods to do real testing (not mocks or unit tests), it will get done with time and money to spare.
> Just because you use AI doesn't mean your brain grows a black hole from which information can never return.
How do you know this? I’m not taking that risk
Every single second of our existence is taking risks about things we don't know. Of course he doesn't know this, he just assigns odds, as we do. Clearly you've assigned much higher odds.
12 years ago I tried to make a simple app for myself. It would display bars that got smaller as the day/week/month got shorter, and would show the weather as a set of bars between max temp and min, cloud cover, etc.
I got it working well enough to display what I wanted in text and ascii, but I could never get the interface good enough to want to use it daily, and certainly couldn't get the graphical interface working. I threw it a Claude Code, told it what I wanted the graphical interface to look like, and let it run.
It got an app exactly what I wanted, and even found a bug in the date parser that I hadn't noticed. I now have it running in the corner of my screen at all times.
The next app I'm going to build is an iPhone app that turns off all my morning alarms when the kids' don't have school. Something I've wanted forever, but never could build because I know nothing about making iPhone apps and don't have time to learn (because of the aforementioned children).
Claude Code is brilliant for personal apps. The code quality doesn't really matter, so you can just take what it gives you and use it.
> Claude Code is brilliant for personal apps.
Agreed.
The clipboard manager I had been using on my Macs for many years started flaking out after an OS update. The similar apps in the App Store didn’t seem to have the functionality I was looking for. So inspired by a Simon Willison blog post [1] about vibe coding SwiftUI apps, I had Claude Code create one for me. It took a few iterations to get it working, but it is now living in the menu bar of my Mac, doing everything I wanted and more.
Particularly enlightening to me was the result of my asking CC for suggestions for additional features. It gave me a long list of ideas I hadn’t considered, I chose the ones I wanted, and it implemented them.
Two days ago, I decided I wanted a dedicated markdown editor for my own use—something like the new markdown editing component in LibreOffice [2] but smaller and lighter. I asked the new GPT 5.5 to prepare an outline of such a program, and I had CC implement it. After two vibe coding sessions, I now have a lightweight native Mac app that does nearly everything I want: open and create markdown files, edit them in a word-processing-like environment, and save them with canonical markdown formatting. It doesn’t handle markdown tables yet; I’ll try to get CC to implement that feature later today.
[1] https://simonwillison.net/2026/Mar/27/vibe-coding-swiftui/
[2] https://news.ycombinator.com/item?id=47298885
Could you share the source to your Markdown editor? I'm always looking for new ones
Drop me an email; I’ll be happy share the code privately. My contact information is on my website, which is linked from my HN profile.
Absolutely. I love building things, but sometimes I want something built. LLM assistance is great for when I want a personal tool, code quality be damned, for a specific purpose, without it taking over a weekend.
You don’t need to build an app. You can use the built in Shortcuts app.
create a shortcut that turns off all alarms. Can have it read your calendar or whatever as signal to determine if alarms should be on/off for a certain day/time and have it run at a regular schedule.
I could, but what's the fun in that!?
(But in seriousness, I hadn't considered using shortcuts. It's not clear it's extensible enough to do exactly what I want, but I'll look into it)
It’s tedious but likely possible.
If you really want to engage an LLM to help point it towards Cherri (https://github.com/electrikmilk/cherri) to help with implementation
They’re way more powerful than you except. I’ve recently rediscovered them and I really couldn’t find a use case for a custom iOS app that wasn’t covered by them.
Where's the fun in purchasing an app from Anthropic?
Co-working with AI is an important skill to learn these days. Similar to paying a bit for AWS for your personal projects as a good way to learn all the AWS tools for your career.
What is the skill that needs to be learned? I've been forced to vibe code everything at work, there's no skill required to ask Claude code to do something.
I think there's a difference in using claude code at work to resolve issues or user stories which are patching existing software and already define what is trying to be solved and what the acceptance criteria is versus using claude code to build something from scratch, where you are acting as an architect.
It leaves more room for skill expression when you're making architectural decisions, defining scope, and designing the application.
Keep telling yourself that if it makes you feel better
I've found that adapting my thinking to how LLMs work best is a real friction point. If you're not doing that, it spits out junk. Your job just has low standards, get used to it.
Yes, realizing the fact that most jobs have low standards, and adapting to LLMs being good enough for these low standards is certainly a friction point. Giving up caring is hard.
Meh. That's fine if you really don't want to build things, and are mainly concerned about increasing your market value.
If you like creating, buying software from Anthropic is boring as hell.
But I thought LLMs democratize development?! Now AI is a “skill” that you have to pay for? Shocking!
Or, you know, you can just build an app. It's far easier and faster to make such an app with claude today than learning a so-called no-code tool.
When it comes to side projects, most of the time, if the spirit isn't willing I find it not worth doing. Process/experience over results, I call leisure. Results over process, I call work. If you have many side projects done mainly for the results, than you are working in your free time, and looking at it like that: is it really free? The modern age already requires of us more results than the spirit is good for. I like to leave side projects for the good of the spirit. An exception could be results for a greater good that one really believes in. This can give purpose and enrich the process and experience of doing.
I have a lot of hobbies. Programming is one of them, but not the only one. There are times that some piece of software would help me with one of my hobbies, but I don't want to steal time from hobby X to build the software. And often these don't involve the kind of coding that I want to be doing for fun. This has been a sweet spot for LLM-aided coding for me. I've built several hobby helper apps where the goal was making one of my other hobbies more fun rather than programming. It's still hobby time, not work, but the hobby is not coding.
If you're coding for the sake of coding, maybe. If you have itches to scratch and ambitions, but can't summon the motivation or the time, then how is that "working in your free time"? A project that used to take up my entire weekend or vacation can now be knocked up in 15 minutes. That's the exact opposite of working.
What is coding for the sake of coding? I don't think anyone does that. Its about solving puzzles, using your brain, learning by doing, creating things- none of that happens when you use llm coding tools. Instead all you're doing is creating more cheap mediocre throwaway crap just because you can.
> What is coding for the sake of coding? I don't think anyone does that. Its about solving puzzles, using your brain, learning by doing, creating things- none of that happens when you use llm coding tools.
Why do you think that? I do regular ol' coding at my day job and have been vibe coding some side projects. They both require using my brain and both require my input for something to be created. They are different, though.
> Instead all you're doing is creating more cheap mediocre throwaway crap just because you can.
It probably is these things but since I'm just building stuff for myself, it hardly matters.
I've written a lot of code and a lot of that has been doing roughly the same thing. It's not a mental challenge; it's a chore. Sometimes it is really gratifying to code and try to figure stuff out. Often times it is not. So when it comes to building something in my free time, I'd prefer to avoid that sort of mental friction and banal tasks just to start working on the actual problem. More so than that, I'm building tools for myself to make my life easier so I can spend it more on something else.
I ride an electric bike with pedal assist. Does that mean I'm not really bicycling? Some might say yes and that it defeats the point. To me it ensures that I pick the bicycle more because it reduces friction to do so. I know that if I encounter a hill that the pedal assist will help me up it and thus I use it more and the net benefit outweighs the downsides. I think it's the same thing here.
I don't take pride in the work that an LLM does for me but I will happily benefit from it. It's a tool.
Bicycle is a surprisingly good analogy. If you cycle for the exercise or the challenge, an ebike sucks. If you cycle to get from point A to point B and be healthier than a car, it's great.
And just like with bikes, people who take pride in doing things the hard way can continue to do so. And they shouldn't belittle people who choose to use assistance.
I share this view, I think it's very healthy.
I've been programming for 30+ years now, but I've always been fine with command line applications. Only recently I started getting into Qt to add a UI and turn my stuff into a real desktop application. It's been a real steep learning curve but I'm finally over it more or less.
Anyway I posted a screenshot of my application on LinkedIn, and mentioned it would be free and open source. I got HUNDREDS of comments from "LinkedIn-type people" all big name engineers that wouldn't HIRE me for anything but either made comments like "looking forward to integrating this into our workflows" or "not the first time someone tried to do this..."
Either way, instead of feeling motivated, I got the worst feeling that I'm doing all this work and people are either going to just take advantage of it and get the credit for "finding" it, or criticize it simply because it's not for them.
It bummed me out so bad that I stopped work on it entirely for like a month.
Anyway I finally came to look at it the way you mentioned. What I LIKED was the process of learning Qt and seeing my old programs come alive.
So instead, it's my "project car" now. I build it up and tear it down all the time. Totally redesign the data models just to see what advantages different designs can give me. Try make my own graphical views. Try implementing language translations.
It's been "finished" for a while now but I probably have five completely different-under-the-hood versions of it and THAT is what has been fun.
I use it constantly all day at work and I never mentioned it on LinkedIn again lol
I've been doing the same thing with my CLI framework. Rethinking and rebuilding with no pressure to finish is a lot of fun.
I've been using the CHANGELOG [0] as a "blog" explaining WHY I made the changes and that's really been fun to look back and read too
0: https://github.com/bbkane/warg/blob/master/CHANGELOG.md
Three notetaking app attempts sitting in my private repos, all stalled at the gap between idea and free time. With Claude Code I finally got the one I really wanted out in two months. Building it has been the best hobby I've found. Beats games or scrolling. When you've been carrying an idea for years, the app that finally ships has more of you in it, and I'd bet we'll see a lot more of these from solo builders.
But who will buy it? No knocks against building old projects, but the market will be flooded with extremely speciality projects. I miss when every app had a spec on the box. I think we need something like that for usage. A new modeling language or something.
>the market will be flooded with extremely speciality projects
All the personal tools described in this thread are duct tape and bubblegum under the hood and nowhere near productionizable. That's what Claude Code makes for you.
The whole point is that for personal tools, code quality never really mattered since it's never going to be exposed to the public or be iterated upon by a revolving door team of devs like real software products. These are all highly overfitted tools that shave off like 15 seconds of time in the day for some particular person.
It's almost exactly like having a 3D printer for software, with exactly the kind of quality that a present-day 3D printer gives you.
I really like the 3d printer analogy. You can still make some pretty cool stuff, you can make some pretty complex things if you carefully design the whole system and put in the effort to print each part individually, and quality depends both on how good of a 3d printer you have and on the proper use of it. "3d print me a new house" is still a pipe-dream: you'll get some miniature facsimile of a house, sure, but a proper house requires proper tools and expertise.
Who cares? Nothing wrong with trying to make a product to sell, but projects dont have to be to sell. I've been having a blast lately working on an old game engine I started during covid and getting sidetracked into some new projects. None of them will ever make me a dime but I'm learning a ton and having fun.
Selling would be nice, but a lot of solo builders are mainly in it for the building. The note app I wanted didn't exist anywhere. I was close to hiring a contractor to build it, so it's already worth tens of thousands of dollars to me. I want to make a bit off it but not run a real business. No VC, no employees. If it earns enough to pay my own salary, I get to keep building it as my job, which beats most startup exit math. Reaching maybe 100 people with the same problem is the target.
Do you have a link to the note app? Just curious what could make it unique, not criticizing anyone.
Honestly, "note app" is a stretch. What I wanted was a knowledge system modeled on PoIC, a Japanese index-card method that predated Evernote. I tried building it on Obsidian and other markdown tools for years and structured notes never stuck for me. I just wanted to talk into my phone and dump quick fragments, then have something extract knowledge and tasks out of them. LLMs and Claude Code closed most of the gap. Now journals, web clips, and daily deep-think sessions all land in it, and Claude Code semi-solves the tasks it pulls out. I'm probably ADHD and it seems to hit for that crowd. Most people would look at it and say "why not Obsidian" lol.
OSS, with the skill + CLI layer built around Claude Code: https://github.com/rillmd/rill
Mac app at https://rill.md/. Still very much in development. If you give it a spin, I'd love feedback either as a GitHub issue or right here.
I've always had more ideas than I can take on. Some of them are good ideas even. With AI tools, I'm now able to generate fairly decent working things for more of them.
Ironically the value of implementing these ideas is dropping fast. A few weeks ago, I built a little search library that runs in the browser and doesn't need a server. It's styled after Elasticsearch and has most of it's term and matching query support, aggregation support, and I added ANN vector search as well (uses web GPU). Most of that was just me going "let's add feature X" and boom done. I used it in some websites (also built using AI) at this point. It doesn't scale but it's great for blogs, documentation sites (https://querylight.tryformation.com/, this site documents the library), etc. It all works exactly like I imagined it would. I probably could add most of the long tail of features Elasticsearch has to this library with very little effort.
But the flip side is that the library got a rather lukewarm reception on Github. It seems people are too busy coding things themselves with AI to appreciate other people's efforts much. And fair enough, if you need a search library, you could probably generate your own. Or just let the AI pick one for you. It's not like this was hard for me or a lot of work.
The economic value of these projects is dropping rapidly. I still like doing them because I like building stuff. And I think there is as a learning curve with these tools that is important to master. Because there is a lot of work that is going to need doing still and people will pay less for it and still expect decent results that you can only get if you master the tools. The ambition level will just go up to match what is now possible. People thinking that they are going to lean back while the AI works for them are in for a surprise. I work very long days the last months.
In the time I’ve had agents I’ve never abandoned more projects. Vibe coding especially just leads me to feel no attachment. I don’t feel proud to put my name on it.
Despite coding from a young age I always thought that I cared more about the outcome than the code. Turns out that’s not entirely the case.
My experience matched this almost too closely. Had a comparison-aggregator side project sitting at maybe 20% done for over a year. I'd open it once a month, scroll through the still-need-to-do list, get tired, close the tab. Pairing with Claude for a couple of weekends got me past the half-built wall. The thing that surprised me wasn't raw speed. It was that I stopped having to spend an hour reloading my own old code into my head before I could do anything. That re-entry tax is what always killed it for me. The hype is annoying but the people dunking on these tools mostly aren't using them for the boring stuff they're actually decent at.
I have revamped my home networking setup, installed raspberry pi, setup adguard on it for the home network, migrated my jellyfin set-up to it, integrated all these devices on tailscale, dealt with physical LAN point limitations by setting up a switch, repurposed old router to get around ISP imposed limitations, setup a three tier backup solution and so on. In themselves they are not particularly challenging but taken together, these personal projects had been pending for years.
Not any more. Few weekends and everything has been implemented and I have learned a ton in the process. It has been great fun doing this turbo charged tinkering. I think personal projects are where LLMs shine the brightest.
Same arc here — Proxmox host with 5 Linux VMs, 2 Windows boxes, a Pi, all of it the "I'll get to it next weekend" pile that sat for two years. The thing that broke the dam wasn't the coding speed, it was no longer having to write the boring middle layer myself: apt upgrades across 6 hosts, auditing which containers had restart=unless-stopped, certificate rotation. I describe the goal, it does the SSH dance. Whether that's net good for my skills long-term I genuinely don't know yet.
I made it document it all too. So now that is effectively a skill which means next time I don't need to give it background on my home networking setup. If I want to check something myself, I can read that same document which is part of my Obsidian vault that gets synced to my phone via Syncthing. So I can basically connect to the pi via termux on my android and check stuff from anywhere.
Plus, time flies on train rides when you are watching a movie of your choice from jellyfin running on the pi. Very satisfying.
> In my mind there are different buckets for personal projects. One is things I do to learn and grow and the other is things I really wish existed.
Pretty much 100% of projects I've done with vibe coding/engineering is in the second category. Stuff I need that either doesn't exist or exists, but is either horribly complex to configure or is a mess of 420 features even though I just need one of them.
It's a lot easier for me to implement that one specific feature just for myself than keep vigilant on an existing app's eventual scope creep as it progresses to the eventual ability to read email[0] =).
[0] http://www.catb.org/jargon/html/Z/Zawinskis-Law.html
I'm at peace with unfinished projects. My dev/ is a graveyard, only a few of those eventually make their way to a public forge. That's okay. Not every project needs to be finished; not every project needs to be published.
I had fun and learned something building those prototypes. I suppose the end result wasn't really the point. Throwing LLMs at doing the thing deprives me of that joy, kills most of the learning opportunities, and leaves you with a (usually half) baked product that you don't really own or understand.
Sometimes I just want a quick bespoke tool and using an LLM is fine. But I do not enjoy using something I don't understand.
With AI coding I was able to build three applications I always wanted but never had time to code them.
Now it is different in a way — I don’t have time to use them.
I've genuinely lost count of the number of little vibe coded things I've built but then failed to use, because it turns out I have limited bandwidth in terms of fully trying out the quirky ideas I'm popping out through coding agents.
I work on enterprise SaaS applications - people forget about features all the time and make requests for things that are already there.
It is more of a bandwidth thing than need of actual new features. When on daily basis you have 10-20 apps you use for work.
There certainly is some relaxing value in working on projects to vibe code them; but not enough to pay some random corporation. Get yourself a Mac Studio or AMD395+ and pi or opencode, and a few plugins and they're pretty capable. Since they're not speed demons but reliable compaions who are always there, you don't ever feel compelled to constantly attend to whatever they're doing.
And when you inevitably get bored with it, well, you've not done much anyway. You can always get back up to speed in a month and have the LLM remind you of what it was doing.
> And when you inevitably get bored with it, well, you've not done much anyway.
I'm very interested in Local LLMs but the cheapest Mac Studio right now is more expensive than 8 years of a Claude Code Pro subscription, and incomparably slower/less capable. If I get bored with it, I will have a piece of unused hardware and a couple grand less in my bank account.
I have opencode with qwen 3.6 on my local machine. Just get the setup right and it's surprisingly fun to work with.
I had a ton of fun setting up and trying it out locally (also opencode and one of the qwens.) I still don't have hardware powerful enough to feel like it's meaningfully productive, but all the learning I had to do (and all the bonus things I got curious about as the curtain peeled back) got my nerd brain all worked up, and finally seeing it work was exciting in that cool-new-experience way you don't often get to enjoy :)
Yeah this is exactly how I felt! Never really felt excited about llms or agentic workflows before. Getting everything setup 100% local and tweaking it to exactly what I want and having it actually working quite well has been a really cool experience.
AMD 395+ w/128gb is all you need. the idea that mac studio is the default is a nerdfest.
I admittedly haven't done a ton of research lately on AI capable PC hardware because of how nuts prices are right now, so I might be missing something...
...but all the AMD 395+ machines I can find are even more expensive than the aforementioned cheapest Mac Studio. Mac Studio starts at $2,000 (only 32GB), AMD 395+ 128GB machines seem to start at $3,000 from what I can see.
the QWEN-3.5-CODER-NEXT fits in half the 128GB and the rest for context. with the right plugins, particularly context pruning, ive got it running over night by writing plans then implementing.
i do not know if theres a smaller model with same capability, but model size and context window at 128 seems like a sweet spot.
token speed really isnt a bother because im either just multitasking or working on the filling in the missing details.
regardless, i think comparing first VRAM sizes w/target model then speed for your cost efficiency. plus, a healthy skepticism of mac hardware costs.
If you already have a gaming pc, then it's worth exploring as the cost of boredom is negligible.
I did tinker a lil with mine! RTX3080 with 10GB VRAM, 5600x with 64GB DDR4 - not very good but it was very fun and exciting to tinker with :)
My partner on the otherhand has an M3 Max 64GB which I've had way more success with. Setting up opencode and doing a tiny spec-driven Rust project and watching it kiiinda work was extraordinarily exciting!
I find $200/month for the pro/max subscriptions cost prohibivitve, but as a software enginere $20/month is just lunch.
And with a Claude or GPT $20 Subscription, i can do other fun things too like using it for real things (emails) or image generation.
A Mac Studio or AMD395 is neither of it. And its not just a basic setup either. I need to buy it, configure it, put it somewhere. That alone is a grand and more + a whole weekend.
You need to factor in the constant value proposition that cloud providers will absolutely drive you to in the next 2 years; even if you're not an AI hater, you should listen to ed zitron's description of the value props these clouds require to make a profit for their VC backers.
This means oyu may be opinionated today on something you will not have tomorrow, 6 months, a year. All that work flow you salivate over can be ripped away.
If you're fine with that, and you've "escaped the permanent underclass" congrats, this opinion is not for you.
I will be able to setup when necessary but until then, your proposed hardware is still very expensive and cost prohibitive to a lot of people.
And at least in germany, running it yourself with free hardware costs me also 20 $.
I'm happily playing around and using open source models, i do not mind at all.
Btw. Open Router also hosts/provides open source models. That might be the better suggestion to the avg reader/user.
Buying hardware is paying a "random corporation". Make the massive hardware purchase after finding out if you have enough demand to buy rather than rent,
My hardware won't be nerfed because a cloud business requires sacrifices.
Yeah. I setup opencode + qwen 3.6 last weekend.
It's actually really cool to have it work on some internal tooling and stuff while I work on my primary projects.
I'm surprised how easy it is to setup and that it can handle modestly complex planning and development flows.
[dead]
> things I really wish existed.
Just last week I was looking for a way to move all the windows from one screen to another in a go. After evaluating many clumsy or over-complicated existing solutions, I asked copilot to write a C program to do it. It had to be minimal and not depend on any runtime framework. A few loops later I had what I wanted without installing third-party tools!
> move all the windows from one screen to another in a go
Sounds like a job for 10 line AutoHotkey script.
In fact it was one of the suggested tools. But installing a whole new application, and having it running in the background all the time listening for evey keypress, just to do its unique job once in a while... well, that's what I was referring to when I mentioned "over-complicated solutions".
Of course if you already have AutoHotKey running for other reasons, it fits perfectly.
I once had a project for novice freelancers. The idea was to take your abandoned fumble and put multiple people on it at 3rd world rates. There are lots of people in cheap coutries who can somewhat code. You can aford hiring someone for [say] 3 euro per day. If they describe where they got stuck in the daily progres report you can look at it and help a bit or not and Just let them figure it out.
You can pay more of course, buy them a computer, an internet connecties, books, courses, even an office but it isnt required.
Just pay 60 per project every 4 weeks and ignore it. If interesting progress happens its fun to look at.
This sounds similar to how art patronage used to (still does?) work.
Which site did you use?
I built a DNS MX health check , including the charting, scheduler, alerts, notifier in just a few minutes . And it’s way more fun to own all of the components. No third party deps, either. I have about 20 random utilities like this that each would have required a subscription .
It’s fun to try to optimize the tool to the extreme. Can I get this to run in a few kb ? A docker image < 100kb ? What’s the smallest possible browser tab ? (sadly it’s still 25mb) . How many ns can I get this operation?
I enjoy building tools my way . Instead of depending on mySQL or a bloated web framework just to get the functionality I need, I can design all of the requirements according to my preferences.
I hope AI actually rekindles the ingenuity of 90s coding. No more “best practices” , “laws of software engineering” or other suffocating constraints. It became way too conformist over the past 15 years ( around the time Clean Code / Design Patterns became a religion)
Not just never going to finish in some cases you never even got to start on them.
You have been architecting this thing for 15 years you know the entire stack inside out at this point, so much so that when you finally say let me see what the AI can do, and you braindump into a markdown file and have Claude build from that spec you are astounded.
Every time I see a story like this I like to play the game "does this person just happen to work for a company that sells AI solutions?". And yes, yes they do. Almost never will see you a story promoting AI solutions from someone that isn't directly involved in selling AI solutions.
+1 and - The title has that weird corpo media "here is why you should not quiet quit" vibe to it, doesn'it?
“Losing GPT 5.5 feels like amputating a limb”, - unnamed NVidia engineer.
The tooling landscape has changed so much in the last two years that I find myself re-evaluating automation setups that were solid 18 months ago. The time investment to rebuild is real but the efficiency gains on the other side are worth it for anything you're doing more than a few times a week.
We have closed-source and open-source software, I think the next phase is going to be self-source or llm-source software (not sure what the term will ultimately end up being), but basically if you have a need for something that is not filled exactly by any app, you will just give the spec to an LLM and in X amount of minutes/hours you will have bespoke software built just for you to use personally, fully tested and to spec. For example, I still haven't found a workout/weight lifting tracking app that does 100% of what I want and at this point I may just build it myself because I could probably do it in half a day with claude and won't have to pay any annoying subscription fees. Maybe it'll still be called AI slop but if it works it works.
>I initially drafted this before my last post on how Claude Code is getting worse. I'm putting it out now so I can reference it in a future post on OpenCode.
IMO, what is getting worse is not Claude Code, the CLI tool, but Anthropic API. That's what most people experience.
I used Claude Code with GLM 5.1, MiniMax M2.7, Kimi K2.6 and had pretty good results.
I prefer Claude Code over OpenCode because most plugins and skills have best compatibility with Claude Code.
Even Terminal Bench showed a bit better results for Claude Code than for OpenCode.
I was and remain skeptical. But I can't help but see how much productive I am and how much fun I'm having using Claude. The landscape is changing.
it's a top affordance of AI because pause development usually over complexity I can't manage or knowledge gap that I can't close in time to sustain momentum. Not only are those roadblocks a non-issue, my perspective shifts from in-the-weeds hacking to a perch with meta view to take the project to completion or maybe beyond previous goals. win-win to the power of win
Honestly, LLMs have been quite good at helping me procrastinate less on my pet projects.
It took me a weekend to build a little esp8266 gadget that automatically turns off my parents' water heater after a fixed time. They saw a dip in our power bill thanks to that. I did not want the complexity of a full blown home assistant setup there. And the Philips Wiz socket did not support setting timers/simple automations.I wanted to test out micropython too for this instead of the usual C++/Arduino/etc..
I started on a saturday and went from 0 to 90 that single day. The product I had after that day was mostly complete but it was crashing at odd times. I spent another day to get to the root cause of the issue - caused by the LLM mixing up async and sync functions. One prompt to the LLM and the problem is gone. The product has been running flawlessly since then - for months.
Every weekend I spent with Antigravity - I was able to push out something totally usable. Water Heater monitor, configurations for the custom Tuya devices I installed in my home, a chrome extension I wanted to use at work.
I am honestly impressed at the pace at which this whole thing improved too. Just 2 years ago - it was a totally useless tool that couldn't even match the speed or accuracy of my text editor snippets.
For me, the journey is more important than the destination. Using coding assistance takes all the fun out of it, as well as the learning. (Not to mention adding aggravation of dealing with prompts and figuring out whether the response is correct or not).
I love to use LLM to kill/archive old projects. I had some gems that i wrote back in the days, checked this week if its still relevant, if rails picked it up already, etc.. then i could just update README.md and archive them without any fear.
I prompted Claude code from my phone to revive Draw!, a pixel art editor I first published 15 years ago as an exploration using Redis as a DB.
The new version is live at http://pixel.drawbang.com after 1 week of prompting Opus 4.6 with a max subscription.
Could you try to revive it by hand? I’m curious as to how long it would have taken.
Absolutely insane request.
Not the whole thing. Just a day looking at it. Did they do that?
I've had several product ideas a week for years, all of them just sitting in Apple Notes. In the last month, I built three of them into a usable beta and I use all of them every day.
I'm using them just to start even more of those projects. I guess i'll always be, most literally, a beginner.
In 2020 I was in the Codemirror forums trying to get help with a project that replaces React in GraphiQL for Svelte. It was too difficult for me to figure out so I shelved it. Yesterday I asked Claude to make this happen and it referenced that very thread I made.
Anyhoo, I'm working on making it pretty (it works!!) before integrating it into my opinionated GraphQL server[1].
There really is no excuse for NOT being the change you wish to see in the world anymore.
---
[1]: https://code.webb.page/eol/gq.git/about
I’ve had a similar experience.
Having better tools really makes a difference when revisiting old or half-finished projects.
So, what are you planning to do when you finish all your unfinished hobby projects?
This is one of the best uses of ai and a lot of fun.
projects you were never going to finish should stay projects that are never finished :)
This. It is bizarre to read about +100 or whatever generated tools. Reminds me of those “1000+ games” bundle that are cheap knock off of popular games.
effort needed used to be a gatekeeper for bad ideas
now Claude will gas you up and tell you your bad ideas are actually the most amazing thing it’s ever heard
tangent: Can someone implement a way to run iOS apps on linux?
I wouldn’t say it’s impossible but you have private APIs, binary incompatibility to overcome. but the AI will eventually conquer both.
Sounds a lot like that disgusti g corporate press propaganda tbh, of the "eat less avocado toast if you cant afford rent" variant. Is the AI mainstream that desperate in their relentlesss push for adoption of their bullshit text generators?
Yes, you can revive it into worthless slop. Learning next to nothing for yourself and achieving a hollow sense of satisfaction at best.
Maybe even shamelessly post it as a Show HN along with the other 99% of worthless slop submissions there.
I want app X. I know how to code it but it'll take a whole weekend/vacation days. I would learn nothing. Now I can use AI to make it. What exactly am I losing here?
How do you know you'd learn nothing from it?
About three decades of programming experience.
Your dignity.
As you perch comfortably in your Throne of Purity, it's worth remembering that the average software developer sits on top of 10+ layers of abstraction (not counting hardware).
LLMs are not a level of abstraction, can we let the meme die already? For the same reason why literature summaries are not an abstraction over reading.
The generated code is an abstraction, as you do not need to concern yourself with how it was made, only the surface API. Saying "can we let the meme die already" is not an argument.
It might be an abstraction, but definitely not the same level of frameworks, OS, machine code, etc.
LLM is not a runtime. It might be something akin to non deterministic compiler where it converts your MD to code.
But why give Anthropic/openai our money? Nonsense. Use open models
Quality, simplicity, speed.
I have a ML Setup with 2 4090 and 128gb of ram, its warm when i use them for finetuning or batch processes.
I do not run them for coding. Its a lot easier and nicer to play around with better models for just 20 $.
Well they are subsidizing us for starters.
The author got $50 free credits.
Also Anthropic is by far the best, open (local) models are glorified autocomplete at best unless you casually have 20k€ worth of hardware at home.
Disagree. Qwen 3.6 and opencode have built and helped plan entire feature sets such as vectorizing and searching, setting up UI to manage categorized search data. Some test systems around this, etc.
Very usable locally assuming you setup your local tooling correctly and you are an actual programmer who can generally help drive this stuff correctly and not just a vibe coder.
How big of a Qwen model are you running that can plan and implement entire feature sets?
I’ve tried multiple that I can run locally and they’re all very much just glorified autocomplete, but slower - on a M4 Max MacBook
I'm running qwen 3.6 35b.
I'm using opencode here's one of the projects I've had it complete - just so you know exactly what it's getting done.
I have a large (300,000k loc) sims-like game that I've hand written over last 3 years.
I have a lot of internal administration tooling that has to be built to manage stuff like icons, NPC brain data, world lore, world actions, all kinds of 3d game data, etc, etc.
One example I had qwen do: Work with me to plan out a feature for an admin panel to manage searchable vector embeddings for each NPC's personality, this was around 600 loc across 4 files, back-end database, front-end UI logic, and front-end templating.
It made 3 small mistakes I told it to sort out and fix, which it did.
I essentially let it do it's thing while I was working on main game core coding. So I was pretty hands off and it planned things out nicely before-hand and got my approvals before it built it.
I really wouldn't call it "glorified auto complete"
Thanks, I really need to try that today - although with pi.dev =)
I do genuinely think that the future is in local models, the online stuff is 100% VC-powered cash market share grabs that will start failing when the forever loop of billions gets disrupted enough.
But at the same time my attemps at comparing claude code + any local model have been such a clear win towards claude I can't bring myself to use a local model seriously just out of ideology.
Yeah I totally agree on the future of local vs. the VC powered markets.
I also agree the reality is that anthropic/sota models are much faster and much smarter, so if you just want to move fast and have them build for you - I get that these local slow models wouldn't be ideal.
For me, as I said I have a large, super complex primary project that overloads me cognitively, and then backend admin dashboards that are relatively simple and isolated/modular. Just due to this specific project, local/slow models are fine as I just check in every 15-30 mins or whatever and answer any questions they have while I'm focusing on the main project. So basically I just happen to have an ideal scenario/use-case for local models right now.
Oh and I did have to mess around with model settings and stuff to get things to work well. I also started with Cline which sucked badly, and then open code actually had 3 major bugs 5 days ago where it was almost useless in large code bases (e.g. freezing and locking up your project constantly), but they ship multiple updates every day, and those bugs have been resolved. So, it's all definitely moving fast and more for side projects or hobbies rather than production I'd say. Still, I'm quite excited with the progress from a year ago and super hopeful about where it's all headed!
Why assume local when you can easily use any of the open models via openrouter or any number of similar services.
The OP said “ But why give Anthropic/openai our money? Nonsense. Use open models”
Then I’d be giving money to openrouter and a Chinese model provider, is that better?
Yes, it is better. They are releasing open models, unlike Anthropic. Additionally other (non-chinese) companies run the open models, so if that is the issue you have options.
"If you don't pay for a product, you are the product". That's always been emphasized on HN threads related to FB, Google, etc.
Are LLMs different?
This is more a form of commoditizing your compliment as they don't see the model as a product but more like a programming language or other tool that you build products with.
Why give apple/nvidia your money?
Jokes on you, i'm running my local models on AMD :-P
[dead]
[dead]
[dead]
[dead]
[dead]
[dead]
[flagged]
[dead]
[dead]
[dead]
[dead]
[dead]
[dead]
[dead]
[stub for offtopicness]
[we've hopefully deprovokified the title now]
To use agentic what? Off topic as heck but I really dislike this trend of coercing adjectives into true nominals - we're using programmatic! - like some sort of even-more-obnoxious variant on the verb to noun ('the ask') process.
Why does it bother me so? I have no idea.
blame the hn title rules (although i would just have substituted "AI")
i doubt anyone is nouning "agentic" of their own accord (yet)
Just "leaders" at consulting firms mostly
It’s ok to use coding assistance tools for anything you’d like! Not that you needed the permission of some random on the internet.
Oh? How very kind of the author to allow me to.
[flagged]
I checked your link and fail to see how that is a pre AI project that you are only completing now, it is obviously a recent concept. Also your username matches the project, so this seems mostly like a lame attempt to shill your product by “joining the conversation”
You replied to a bot.
What is the point of using AI for a side project? Isn’t the point of a side project to A) have fun writing code or B) challenge yourself or C) learn something new, and usually all 3? Slopping out random stuff you have no attachment to, taught you nothing, and that feels bad to pass off as your own work accomplishes nothing.
D) to solve a problem in your personal life? To make an app that you wish existed but didn't?
Most of the replies in the thread that do this just scratch their OCD.
[flagged]
[flagged]
Please see https://news.ycombinator.com/newsguidelines.html#generated and https://news.ycombinator.com/item?id=47340079