Wasting my summer
May 2025:
Discord might just be home to the largest number of live bots, though the actual statistics vary the actual number is definitely in the millions. This summer I decided to boot up my old discord account which I had not touched since 2023.
The first thing I noticed was how largely dependent discord was and still is on these bots for user engagement.
I went through some new features like mee6 ai, or the updated embeds from poketwo which were added to these bots after I left. I just went through some more messages and every once in a while I would see a small embed popup (depending on the bots used) which read user __ has leveled up and is at __ lvls. In retrospect I do now understand that discord was being used at large in late 2022 and early 2023 and bots were a novelty to some new comers like myself at that time. While going through the messages I thought well "that's weird, why are the level up messages popping up after set number of user messages?"
Again retrospectively xp distribution was probably not a huge deal when more prominent issues like mutes and bans were floating around to be coded and they probably didn't have the current resources from js or python libraries like discord.js.
Well turns out that developers or designers ended up using static or formula based xp distribution which seems to have stuck around till date. Recently I was also reading some foundational papers on optimizers like ADAM and momentum, and was also going through the basics behind actual neural networks like weights and biases, activation functions and the like. That's when I thought why not use this to distribute xp and that's how I ended up spending most of my summer.
NeuroLeveL is the output from my two month quest, adapting dynamism in the xp distribution system and devloping inbuilt fairness and spam protection in my bot JASON. JASON for context is the parent bot which was created intially from this idea and is now the larger framework which is published as NeuroLeveL. Other than simply encouraging users to participate, this framework also aims at implicitly discouraging spams and essentially gives diminishing returns to spammers while awarding silent warriors.
How it goes brrr:
In a nutshell NeuroLeveL is an open-source plug and play framework aimed at individualism and dynamism which works by calculating the user activity i.e. how active the server is currently and msg activity i.e. how active has each user been in contributing towards the server activities which it then applies the Perceptronic logic to compute each individual's xp.
Now this means if some has larger contribution to the server activities when there are few users online, that user is awarded a larger chunk of xp, and similarly for someone spamming messages when multiple users are conversing, everyones xp would be lowered. This makes the system pretty robust in modest servers sizes against spamming. Other than this implicit anti-spam, the bot also adds a 1.5s spam timmer to each user for spamming inherently neglecting their contribution and awarding nearly zero xp.
As of this post, JASON is underworks to be posted as a dynamic game environment bot, but NeuroLeveL is open to contribution, if you've read through this semi dev-log/blog and want to check NeuroLeveL out head to: https://github.com/Unknownerror-404/NeuroLeveL.
You should find everything you need, from contributions to improvements there. And to improve the NeuroLeveL framework work on 1.01 is currently underway.
-This was Unknownerror-404.
Top comments (0)