RimWorld 1.7 moves from Unity to Ludeon's in-house Waylith engine | Hacker News Reader

RimWorld 1.7 moves from Unity to Ludeon's in-house Waylith engine

ludeon.com

101 pointssha-356 comments

Notably they were able to, in one update, essentially double performance and cut load times to a quarter of what they were while also maintaining both backwards and forwards compatibility for save files and without breaking 99% of the mods on the workshop. Genuinely impressive feat.

And yet they can't mention or admit that LLMs were involved or the gamers would rise up.

Were they?

Oh noooo definitely not :) Nuh-uh no way no sir.

So, your proof is the absence of proof.

I believe it to be very likely. AI is a development tool that allows you to implement things that simply weren't possible before due to time constraints.

You don't need a plane to get from America to Europe, but there are constraints if you don't use one.

I think most people are angry about the artistic portion like images and dialogue of games being AI generated more than the code. Code can be art as well in my opinion, but it still can be even if you AI just ported a large part of your code base to another target engine for example.

i'm pretty convinced that if you look at the cadence and tempo of updates on Steam among actively developed games that one could probably make pretty good guesses about which groups are using LLMs internally.

Doesn't prove anything, lots of stuff changes development house tempo, but to anyone watching things as a whole updates across steam games in active development are 1) quickening 2) narrowing 3) becoming predictably scheduled ; all of these point to more than just a single group refocusing on the work at hand.

I think builds are getting pushed out a lot quicker, and personally I know of a lot of anecdotes where groups' entire build and output chain are handled by LLMs invoked by cron (or a clone). That alone will add a lot of regularness to things.

This is true in general, but how does that apply to this game? Their release schedule has followed a pattern for years, this release included - a major version every 1-2 years, followed by a flurry of hotfixes that then slow down until the next major release. Wouldn't you see that schedule tighten post-LLMs if they became as AI-pilled as this comment thread wants them to be?

Tynan has been pretty positive about LLMs on his twitter. I obviously don't have any inside information, but I think it would be very surprising to me if there was zero usage of a frontier model in doing something like this.

For a moment there this wasn't about LLMs

LLM use is normalized in just about every other area except in very fringe nerd groups who oppose it like it was an interdimensional threat.

No reason to mention using it any more than disclosing the text editor they used.

Wait, I want to know which text editor they used! Now that you mention it, this is an important factor in purchasing any software. This information needs to be tagged at point of sale.

You're right - after all, just like LLM usage, the text editor you use will have an impact deep in the project and affect the quality of the end product, right?

Judge the project by quality not by the ideological convictions of yourself

Are you telling me every pizza and event place using the same AI pictures and flyers suddenly is already normalized?

Yes if they used notepad over a frontier IDE their refactors may have missed a reference. I demand to know which version of notepad they are using.

Not sure why you are downvoted. Gaming communities absolutely erupt when using AI for building games is mentioned.

Thought that’s more about the art, voice and music. The more artistic side of things.

I think expecting understanding of the nuance of code vs art out of the gaming community might be asking a little too much.

There is a big range of understanding what "AI" even is, even among the most fervent haters. Some people will claim "Any visible AI is bad AI", others will claim "Gen AI is the job stealing one, others are OK" and some will direct their hate towards LLMs, but only running on other's servers.

I don't think you can say clearly there is a one "community" collectively "hating" the same thing here, it's just a lot of people with a lot of different strengths and directions of dislike basically.

Not really? It happens in code too, just most the time its sight unseen unless a team explicitly mentions it.

I mean we've even been seeing gamer outrage over LLMs being used in decompilation projects. How "decomps have lost their soul" now that it doesnt take a small team several years.

I just heard the giantbomb guys describe how to block Claude code on Github so you don't have to play any filthy AI helped decompilations. It's a new kind of social benchmark among some people to be the most against AI.

I don’t like calling ppl stupid but it was a benchmark of “stupidity” if I ever seen one.

Why? Because people with well developed frontal cortex predicted the increase in capabilities easily while people with underdeveloped frontal cortex couldn’t extrapolate today’s improving slop to tomorrow competence.

Hence I have learned a lot about my fellow humans of many political tribes and how the distribution of those more “naturally talented at predicting patterns” look.

I stopped paying heed to certain kind of emotional, outcry driven people that thrive on your attention and take morality as a hostage.

For they aren’t what they want to appear to be usually.

If we all stopped giving them our time they would dry up and wither without audience and the world would be a more pleasant place.

I think that's a bit mean to the people (in my experience most people) who didn't and still don't see it. What's happening right now is really akin to magic. We're throwing incantations at sand and it's doing work for us. The rate of change is such that nobody can keep up with it all, even the people like me who use it daily.

What people DO see is having to use or be monitored by ABYSMAL types of AI at work or when they try to call x company for support or whatever. The outcry against data centers when the same people building them were moaning about how electric vehicles would crash our electric grid completely is quite justified. As is the outcry about creative and intellectual theft. Add deepfakes, scams, spam, misinformation, etc. and the negative externalities of AI for most people seem to far outweigh the "benefits". And everything beneficial seems to just be training the AI to do your job in disguise. Add in that the billionaires behind most of this tech come off as completely out of touch with the real world and I totally see the perspective of "screw AI and everything it stands for" even if I don't share it myself.

I think you are too lenient. Everyone has the same access to the information nowadays. Everyone can ask, even an LLM, about how it works, what does this and this mean.

There are no excuses in this day and age. I am also sometimes lazy but I never pat myself on the back for it.

Why would I hold other people to lower standards than myself? I think that, would be offensive.

I refuse to treat other people as less than myself. If I can investigate this topic without experience or phd then I assume anyone can obtain such knowledge before unleashing scathing critique.

Do not defend people who hate on things without due diligence and who generally are quick to hate.

My guess is there's a whole plethora of important information you're ignorant about that you could just look up and have full knowledge of if you cared. But nobody has time to care about everything. People only have so much time in a day. After 8 hours of working and 2 hours commuting because nobody can afford to live in a city anymore people just want to eat dinner, relax for a bit, and go to bed. Maybe spend time with their kids instead of sitting around reading about how AI is going to make them obsolete.

Example I would give is I'm totally ignorant of local politics. I just don't have any time at all to parse it, and I probably wouldn't bother even if I were magically given the time. There's people out there that would be just as upset at me as you are at people ignorant about AI. And it's not that I'm lesser than the person focused on local politics, I just spend my time elsewhere.

Why would they? They likely don't talk about what IDEs they use either.

They appear to have a pretty well-defined modding interface that didn't rely no unity's custom data structures. I'd suspect this project was started some time back, not just a one-shot port. Can't say whether or not any LLM's were used on tested in this project, but I get the feeling they set themselves up for making it portable from the start via some good old fashion thinking.

LLM's are still kinda shit at porting large in-house game projects. They can quickly and easily mangle codebases.

"the audacity of these people, they didn't even mention that they cheated and used emacs ! they sacrificed their pinkies for development time!"

That's an odd thing to say given that there hasn't been any word on if they even used LLMs. Presumably this update has been in the works for some time.

Mod compatibility is remarkable and much appreciated!

I happen to know a bit about the game's guts because I've been working on a RimWorld-based IDE [1] for the last few months. The C# side has grown to 75k lines of production code, much of it depending on Unity/Verse.

So when I saw the Waylith news, my first thought was "well, now it's either a complete rewrite or legacy forever". Migration took one night and introduced only a dozen regressions, mostly in low-level interfaces that no "sane" RimWorld modder would ever touch.

Along the way, these pains and hacks went away:

- Ancient net472 instead of cute net10.0.

- Mono-specific networking workarounds.

- Mouse wheel-event queue stuff.

- Unity font and atlas stuff.

- Semicolon key handling. Don’t ask.

- Unity/Mono WebSocket bugs. Probably gone. I still use a custom implementation; haven’t tested the new built-in yet.

- Harmony is in NuGet now.

Some weird bugs have migrated to the new engine. Like the game freezing forever on Alt+Tab in "Borderless window" mode, on both X11 and Wayland. Fortunately, Unity hacks still work.

Now to the headliner: performance. I haven't measured round-trip terminal latency yet, and loading times (X11 window to interface) actually got worse in my case.

But before the Waylith migration, an idle, focused SlopWorld window in Eco mode (colony sim stopped, some resources unloaded, map hidden, rendering in 60 FPS VSync) ate 50-60% of one CPU thread. After migration it's 15-18%. So roughly 70% less load for free, before any gameplay or mod optimizations.

[1] https://slopworld.space A shameless plug, hopefully useful for the community.

Wow, that's really cool! And rather unexpected (at least from my side, but I'm also not really keeping up with RimWorld news).

And then, at the end of the announcement, it says:

"RimWorld isn’t the only game on Waylith. We’re working on something new! It may be a little while before we can share more details, but we couldn’t resist telling you."

Exciting!

I think they're being a bit too optimistic about mod compatibility - from a cursory look, a lot of Rimworld mods do at least import UnityEngine (mostly for small utilities), so it'd take at least a recompilation. The fact they test with hundreds of mods does suggest they don't plan on leaving modders hanging, though.

They said they implement a lot of the Unity APIs for mods. If there's something like that used by 99% of mods, they would have easily detected it and implement it.

Someone tested the version with like 1500 mods and it launched.

I tested it with over 200 mods and a save from 1.6 and it worked.

the client bitched about the mods having incorrect version numbers, but after re-enabling them and loading the save nothing was broken (any more than it was in 1.6, at least.) -- and it loaded much faster.

I think there was a lot of in-situ handholding for unity API expectations.

They included a compatibility shim that forwards these calls to their engine. The compatibility will depend on how many UnityEngine equivalents they implemented (most mods only use really basic ones) and how accurate they are.

>Older PCs: If your PC is more than 10 years old, 1.7 may not run on it.

surely not AVX-512 (2015), so what exactly? AVX2 (2015)?

Maybe vulkan?

The new engine is Vulkan based which released around 10 years ago if I'm not wrong.

I've long suspected that writing your own game engine can give you much better performance than using something like Unity. And with AI, it's now a lot more doable.

It's a shame they didn't mention the use of AI and how they used it. It's impossible to make such changes so quickly without a swarm of agents.

Mmh, impossible is maybe too strong.

> RimWorld isn’t the only game on Waylith. We’re working on something new! It may be a little while before we can share more details, but we couldn’t resist telling you.

I suspect Waylith is in dev since long time on this new project and they pushed it to make it prod proof before pushing the game further.

A statement about AI could be great, yes.

Why does AI need to be involved in such a migration. The original game was made by one guy. Its a small team with a deep understanding of their project, it seems counter intuitive to outsource this to AI.

i'm very confused why HNers got amnesia about pre-AI times. It was always possible to build software fast if you had expertise and good code that was easy to work with.

It counted as fast then. It would be really slow in comparison today. Sure, whole engine rewrites with compatibility shims were possible to do in months in the past. But let's not pretend it wouldn't be extremely faster today with LLMs.

it's false gains. good code compounds on timescales that are longer than the year it's been since opus 4.5...

I have no idea if that's true anymore in this situation. The benefit of good design and code is that it's flexible where it needs to be to make changes easy and hopefully avoids many changes overall.

If changes become cheap, the same benefits don't apply anymore. Large scale refractors can become background tasks. "Turn this project into ECS style" is one session away. The things we valued in long timescales happened because changes and backtracking were painful and a huge time sink.

Is this site LLM News or Hacker News

80 % of the people on this site have been so permanently ruined by LLM usage, that they can't even imagine how other people might be able to do impressive things just simply by being good at their craft.

It’s very funny to watch people admit they can’t code by insisting that “no one could do X by hand!”

Let's just wait until the truth comes out to see how smug you were

I’m pretty smug

Why do you claim that this change was made "so quickly"?

Nowhere does it say how long they were working on the new engine.

Not impossible that this was under development for a longer time than just one version.

Reply on news.ycombinator.com