Noodleface
A Mod Real Quick
- 40,475
- 20,529
AI: The Rise of the Machines... Or Just a Lot of Overhyped Chatbots?
- Thread starter Sanrith Descartes
- Start date
- 6,407
- 11,094
Fucker
Log Wizard
- 17,443
- 43,262
rhinohelix
Dental Dammer
- 3,642
- 6,206
Reverse engineering continues to reveal secrets. This is going to be some of the best things ever:
[AsmonGold Short: https://www.youtube.com/watch?v=QRjyh9t3Jeo&list=WL&index=147
[AsmonGold Short: https://www.youtube.com/watch?v=QRjyh9t3Jeo&list=WL&index=147
- 1
- 1
Ill post some whenever I get home. I basically gave it dumbass promp this morning to basically just continue implementing systems and features with really no guidance at all.
I haven't kept up with everything its done all day, but last I saw it spawned other characters to test groups and raids, and had moved on to implementing tradeskills
- 6,407
- 11,094
Only mobile game I ever played was Candy Crush (the OG) and the month I got my CC statement and realized I spent money on in game gold I uninstalled that mofo fast. Now I just doomscroll Reddit looking @ cats.
ShakyJake
<Donor>
- 8,872
- 21,682
I mentioned, maybe in this thread I can't remember, that I've been creating BG3 mods. The current one I'm doing is reworking zones, populating them with different enemies. Initially I was manually pointing to specific places on a map to put enemies here, and here, and over here. Then I just said, Claude, just examine the zone and you place enemies and patrols where you think makes the most sense. Fucker did it better than I ever could.Ill post some whenever I get home. I basically gave it dumbass promp this morning to basically just continue implementing systems and features with really no guidance at all.
I haven't kept up with everything its done all day, but last I saw it spawned other characters to test groups and raids, and had moved on to implementing tradeskills
- 1
Hah, its funny you say that because in most situations where Ive messed around with standing an emulator up, it never really felt worthwhile to actually play. Not that Everquest has any surprises at this point, but it sort of ruins it even more if you the exact math and tables behind it all.
In the same way you figured it out, I realized I could really have it completely rework anything and everything and so long as I dont go digging in the db, I wont unless I play it.
- 1
rhinohelix
Dental Dammer
- 3,642
- 6,206
So as I have mentioned previously, I grew up near the TX/LA border, and LA always had a lower drinking age/a more permissive drinking culture than TX, so always held allure for teens/young adults of all ages from both states. Later on as a young adult I ran security in one of the nightclubs in the "hot corner" right at the border, and many of the things and places I found magical as a teen (ahem) where not nearly as amazing when I went there every weekend to work, watching people act like idiots and be responsible for their safety as a 25yr old wearing shin guards (to protect against the weighted bottoms of tables), to see the clubs with the lights on, as the bars were prepped and the DJs set up, it was completely a different experience than it was as a 16yr old sneaking in with a fake ID a little drunk with the light shows going.
So much of the magic of EQ was the mystery of the Old World, the rare spawns, the drop rates, and legends and the myths. I know all these years later we have learned the truth of most/all of them but to look at the bones of the code and see it live from a spawn table/data architecture perspective. It's still pretty neat, though, what they were to make, the magic they were able to create with those tools and tables.
- 2
Control
Avatar of War Slayer
- 6,421
- 17,816
eqemu is basically eq source so you should be able to basically make it do anything. someone posted a little while back (maybe here somewhere) that Eve released their engine code. glorious days for anyone that wants to dump time and tokens into oldschool mmos!
also (mostly) unrelated ai rustle of the day (actually a whole lot of days...):
<ask claude to do literally anything> - "that will take approximately 37 years! slotting for post-release expansion"
"the fuck we are, let's do it now!" <actually takes 7 minutes>
<claude asking me for a bigass list of specs and definitions for something I haven't logged into in 5 years> - "Needs You! estimate: 2 minutes"
<me after 20 minutes of trying to remember where the fuck to even start looking for that> - "scratch whatever that was for, add it to the expansion list"
- 1
rhinohelix
Dental Dammer
- 3,642
- 6,206
Before GPT 6.0 hit, I was uploading my POE2 build to Path of Building and my build and using ChatGPT (I guess is was medium-effort Opus 4.8) to scour the market looking for equipment upgrades. I pride myself on getting good stuff for cheap and I swear it could always find things I couldn't. The models' ability to sift through and process data is unmatched, so don't take it to heart, that's all it's doing with the level building, is looking at maps as ones and zeros, and then figuring out where things go.
I have been watching a ton of game dev videos as I dig into remaking Conquest of the New World. Forget all these "One prompt COD" BS. I am at Milestone 6ish of 16 and on PR#28 as I go through and add things in, change things etc. The tools are going to be there to make EVERYTHING. Some guy has already made a Warcraft 4-alike, after a month of work and a gazillion tokens. (Edit 23.3 billion tokens, actually) Ashen Crown.
The problem with AI allowing so much creativity is going to be once again, finding people you trust to showcase the Gems. It's going to be very similar to the start of the Internet, with the walls coming down the barriers to entry for software creation being removed, just like the huge explosion we have for websites/static content with the Internet and then again with media content creation when bandwidth became plentiful enough, assuming we can keep regulation from keeping the price per token high and governments from picking winners and losers.
- 1
Tuco
I got Tuco'd!
- 51,425
- 94,243
Yes, this is a huge issue.
There a couple problems these AI-coded rip-offs present:
1. The scope of changes in some of these vibe-coded pull requests are so vast you can't expect the open-source community to really understand what's going on.
2. Normally an open-source project is built up over many years and the author's reputation grows over time. These applications can be dropped out of nowhere with no known people backing them.
As an example, normally when a AAA game like GTA6 comes out, the modding community erupts in activity. Many of the people involved already know each other and they mostly port existing tools to the new game. The first person to figure out how to extract models is probably currently involved in GTA5 modding. The tool he publishes for that is likely an extensino of the current tool ( probably GTA Extractor ).
Now? Who knows, people will be vibe coding all kinds of shit. The community will likely settle in on a set of tools and will respond to new ones with, "Why would you create a version of something that already exists?", which is a question as old as time. But you can't trust the tools being made as much as you could in the past, even if you build the source code yourself.
Tuco
I got Tuco'd!
- 51,425
- 94,243
A challenge with EQ is that most of the good stuff is server-side. I'm guessing that Claude knows about the information present in https://www.eqemulator.org/ and similar projects and can replicate that in an arbitrary engine like UE. But it wouldn't be able to do the same thing with, say, the new expansion that's in beta right now (Favors of Fortune).
What would be interesting is an AI that can watch videos of people playing Favors of Fortune and combine that with logs and parses and piece together the server-side information like monster names, monster stats, spawn locations, drop tables and quests.
Either way, this community has a thriving EQ contingent that loves the game (including myself). Keep us up to date on your progress!
I'm also curious if you've considered using Godot as an engine instead of Unreal Engine 5.
Noodleface
A Mod Real Quick
- 40,475
- 20,529
One thing interesting - not specific to EQ, but MnM - is Claude already translated the packets for me. Something it would take a human quite some time to do, it just did in a few minutes. Now, I would probably argue the MnM devs encryption on packets isn't up to snuff like EQ might be, but I could easily make a ShowEQ today in about 30 minutes if I wanted to ruin the game.
I would think you could let Claude monitor you playing in a zone and it would be able to get pretty close to what the server is actually doing.
Rabbit_Games
Ahn'Qiraj Raider
- 3,073
- 6,386
The way I'm currently doing it has the open sourced emu backend completely stood up as if it was just a normal eq emus. From the client perspective I'm using a combination of the trilogy discs, rof2, and live clients that get reverse engineered via IDA/Claude and then ported over to unreal. In most cases, I'm trying to recreate mostly retextured classic functionality.
It's ability to figure out things like the particle system via IDA just blows my mind. I'd have a less than 0 chance of ever being able to do what its doing. Its able to decipher all of its attributes with almost perfect precision.
For example, on live they moved to the new particle system at some point and eventually the og particles coming from the models hands broke (turns out they were still being generated, but were off the map and no where near the character casting them). It fixed that on the live client with MQ within a couple of minutes and just had it working in this project inherently.
It also found og particle functions the game never actually used and implemented that.
And lastly, another thing I forgot was even a thing were bolt spells actually traversed the landscape as an actual bolt. Since I didn't remember that was even a thing, I obviously didnt mention it but it went ahead and added it.
Its able to go way beyond showeq. I have it able to completely re-patch macroquest in about 3 hours post patch. The showeq offsets are rather easy to find, but some structs used in mq are shuffled every patch by dbg in an attempt to make it difficult to patch. The layouts for 2 of the 3 arent ones you can use signatures against to find them post patch, so it uses a combination of runtime scans + best effort of figuring out the datatypes and then comparing the values to the previous patches values.
Tuco
I got Tuco'd!
- 51,425
- 94,243
That's interesting. I'm not familiar with EQ's packet structure, much less MnM. Is their packet structure significantly different?
The way a human would do it (Or at least, what I do when I need to reverse engineer closed source systems...) is record the datastream in wireshark and trigger known events (Like wait for an NPC to spawn) then examine the packets around that time. Do that a few times and you can find the pattern and extract the "Monster Spawned" packet from all the background chatter that exists in any online game like chat messages, nearby combat, acknowledgements, etc. But knowing the event that happened is critical for closing the loop to determine the right packet information. Assuming that Claude didn't just lift the packet info from something it was trained on (Like EQ), how did it determine the translation?
Tuco
I got Tuco'd!
- 51,425
- 94,243
That's funny. Anytime I see the OG particle effects in a video it's a pure nostalgia hit for me, despite having been playing EQ steadily over the last decade. That they haven't enabled it despite people wanting it is just one of many instances of them being incapable.
THJ is just the tip of the iceberg in terms of EQ's development cycle getting BTFO by the community. DPG pays its developers bottom dollar (Like literally bottom 3% of salaries in the area) and they get what they pay for. We're already at the point where random groups of developers in the community can outwork DPG as a hobby and AI is dropping that floor dramatically.