binary0010

I don't think people are talking about the least code possible, just not incredibly verbose and inefficient like what you get by default from llms.

For example I have a game I've been working on for a few years, I do stuff like "implement this simple psuedo physics system to make the bot follow the character like so...etc"

After some planning and back and forth.

It returns mostly working code a little odd on some edge case.

But as I've hand coded this thing for years. I could easily look at it. Laugh my ass off, it had multiple classes and around 1k lines of code, all kinds of crazy non performant crap.

The exact thing I needed, I reprogrammed in around 5 lines of very simple code that did exactly what I needed with no edge case weirdness.

Now the vibe coders actually ship that shit. I like to read vibe code games now and again, and there is no possible way those guys are ever shipping a real game, as every single decision is verbose along with the worst performance decisions over and over everywhere.

Sure it can get you some cute little toy projects, but it will absolutely fall apart if you are trying to make real games.

Don't know about saas apps or whatever. Maybe that stuff doesn't matter at all.

Or you get the original output result was perfect and the adversarial "rethinking" switches to an incorrect result.

What are you going on about?

No idea what you think you are arguing. Are you in the wrong thread?

I said I don't get why people keep posting essentially vibe coded game clones... I get bored checking out GitHub projects on HN that are doing absolutely nothing new and are essentially baby projects made by Claude on a weekend, and terribly architected to boot. Cluttering the feed.

You really don't see where the author uses ai? Pretty much everywhere, all the docs, the overview of the project, a LOT of the code is obviously primarily ai written, etc.

Isn't this the kind of thing you can essentially fully offload to Claude code these days? Don't really get the point of these tiny primarily llm generated game clones tbh.

My younger friends/acquaintences hate ai significantly more than my genx/boomer friends. Could be my own bubble of course... Take from it what you will.

Yeah I deserve to be down voted for the last message no doubt on that lol.

But originally I was just trying to be helpful by quoting their charter on what they consider "agi" now.

I'm so confused why I was down voted for answering the question that was asked?

OpenAI’s mission is to ensure that artificial general intelligence (AGI)—by which we mean highly autonomous systems that outperform humans at most economically valuable work—benefits all of humanity

From: https://openai.com/charter/

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!

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"

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.

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 have opencode with qwen 3.6 on my local machine. Just get the setup right and it's surprisingly fun to work with.

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.

More slop, yay.