wall jumping works now including strafing: https://www.youtube.com/watch?v=6jlG4YCJd0Y
1 week ago: https://www.youtube.com/watch?v=PMtzSYGib00

lol Arie...

I agree with the sentiment about the feeling of fulfilment and being proud of what you accomplished after solving a difficult challenge. What I am curious about is we all have different standards for what we can consider impressive.
I think we can all agree that simply 1-shotting 'anything' is not impressive. But what some people may be able to achieve by writing code themselves or designing something themselves, can seem unachievable for others who lack the technical skill; but after significant input and many rounds of critical thinking, may still feel like an accomplishment even though they have used "shortcuts" by using AI tooling.

Give an example from my day to day job, in my previous roles we were using analyzing data to create new insights to drive new commercial opportunities for CPG companies (Revenue Growth Management)
Classical cycle:
Data -> Visualization & Dashboarding -> Analysis -> Insights -> Decision

First we would clean the data, build all sorts of dashboarding to visualize it in a meaningful way and then use our own knowledge and our own eyes to spot opportunities. We summarize the insights and finally make a decision what to pursue or what not to pursue.

Now, with agentic AI, we are basically skipping step 2 & 3 and going straight from Data -- -- --> Insights -> Decision
You no longer need to visualize the data yourself and we have a full deterministic layer that performs the analysis on an even better way that humans could do. The insights gets presented to you and the human always stays in control of what decisions you make. Always.

Now back to the sentiment of the last few posts, yes it does bring a bigger level of satisfaction when I clean the data myself, build amazing new dashboarding and perform step A->Z analysis myself before reaching the conclusion.
But the importance of this satisfaction coming from performing these painful steps is now being replaced with much faster and qualitative insights.

One example that pops up to my mind is the calculator, or perhaps the typewriter and then computer?
100's of years ago we did mathematics by hand, then came the calculator, then came the graphical calculator and then personal computers. I am sure there has been a huge shift and rejection throughout history when these shifts took place; because the level of satisfcation decreased a lot. I think we all had it in university as well, solving a difficult math problem on paper (many sheets of paper) and then finally reaching the (correct) answer was fucking good. Then came Wolfram-Alpha and people stopped going through the pain already.
Same for writing; a feather and ink vs type writer vs a computer, and now I literally talk to my computer for hours sometime transcribing shit and having a conversation...

Long story, AI is an instrument that can unlock people who are lacking the technical skills / time investment but who do have the creativity in their head and were simply not finding a way to turn these ideas into something tangible, but now they can.
Should every idea turn into something? Absolutely not, see all the massive amount of slop created...

But outsight rejecting certain work because people have used AI assistance because otherwise they could not have achieved it seems like gatekeeping to me. Be critical: yes, absolutely. Be supportive of each other: also yes

Epic

I think it is pretty normal to have a knee-jerk reaction to AI generated/supported content, when most of the stuff you see online are 'one-shots' and barely involved any effort.
It's insane how much content and output is being created the last few months, but I would say the people caring to spend even a bit more effort and treat it as a proper project is less than 1%.

So I get most of the hate but at the same time, why not just ignore it? For those who are not into it - just ignore it and move on.
For those who want to contribute, be very critical and share your thoughts because it helps to improve it.
Over the past few weeks I've had quite a few people (top players even) try out my game and thanks to their feedback I was able to get more ideas and turn them into tangible results.

For me personally it's all about learning new skills and having fun. In my day job I work at a software company but on the sales side, not technical. I am literally training our large CPG clients (Nestle the biggest one for example) how to use our Agentic AI solution to run their business processes better and find more value... so why not try to use my prompting skills and build a whole game?

You can have personal views against AI and personal taste, completely fine, but every single industry is being transformed right now. Finding the right way how to use it in your advantage is going to make a huge difference vs those who are simply resisting it. Video game industry included

https://www.youtube.com/watch?v=PMtzSYGib00

Double jumps in progress

btw congrats to 'ex' first person to beat the bot on the pit (20-18) after losing many rounds :D

Bot can now perform a few rocket jumps on spire: https://www.youtube.com/watch?v=4jRQrMYdxNM

Made further improvements on the bot as well, can't beat him anymore myself in the pit... he is now responding to rockets and actively dodges in opposite direction; switches weapons faster and has no more fire delay (previous 0.9sec vs 0.8sec on average vs human players)

If anyone can beat him on the pit let me know...

New update & improvements, thanks for the input from playtesters!
-> Bot is stronger again on the pit arena, I'm having difficulty beating him... (Other arena's map path w.i.p. see below)
GFX updates:
--> Rockets more visible now. Clearer flame & smoke trail
--> HUD updates: Health cross, weapon icons per selected item, HUD customizer in settings
--> Website UI work in progress
-> Physics big update: Rocket jumps work much better now, closer to tf2
-> Spire arena 1:1 copied from tf2 and first round of clipping cleaned up - still some glitches to be smoothened out
-> Multiplayer bugs fixed (multi-kill on SG & grenades, duel link also working again)
-> Tryhard mode actually working

Still in progress:
-> Prio 1: Re-sweeping all map-paths after rocket jump update (~200 Million flights... 76 Million done as we speak)
-> Bot learned to rocket walljump but not in the map paths yet. Will be added once the 200M flights are done, especially in Cannery looking forward to add wallsjumps.
-> Achievement icons in progress on the background. Done with 10/50
-> Map visuals

So at this stage, looking for feedback about the physics / feel of the rocket jumps... I feel they are a bit more boosted vs TF2 but maybe because I got used to the lower jumps, they feel much better now.
Any feedback on the bot's behavior on Arcadium or on The Pit? I know his rocket jumps are not good yet. Anything else that stands out let me know.

p.s. the different tiers don't impact the bot's difficulty yet. I'm first making a very strong reaper and after that his difficulty will be tuned down to make him easier all the way to bell pepper

https://scoville.gg/

new update, bot improved again, a bit stronger now
-> tryhard mode working better (e.g. if you kill him 3x with LG then he will also mirror that. same for prefire / sg switch)
-> 1st flight paths in for Cannary and Shoreline but not so satisfied yet, Arcadium stays best map to practice vs bot
-> demo reply system working, could already review quite a few demos and make the bot learn from it

Thanks for testing and helping to improve the game!
many more items to be worked on ...

update

www.scoville.gg

Improved the bots brain and playstyle, sniper rifle passive reload is in, mouse sense should feel better now (will enable 1:1 source sense in next release), and important: Every game vs the bot is now recorded and used to train and improve the bot. So the more people can play (Arcadium only at the moment) the better I can improve the bot.

Currently in progress:
- mapping CANNARY flight paths for the bot
- mapping SHORELINE v2 flight paths for the bot
- teaching the bot wall jumping
- tweaking the tryhard mode

WORKING ON IT

including MGE etiquette

added. I saw on your page you played in a mge world championship.
Didn't know they existed, and 2; then I definitely need your input to improve the bot!

added - the bot is far from being better than me but it is definitely my end-goal.

I want to go to a place where Scoville has 2 directions:

1) 'road to reaper' you start fighting a easy bot 'bell pepper' and move your way up through 'jalapeno' etc to the California Reaper -> the first to kill that one gets a prize and recognition, and is placed on global leaderboard + time trial. That could be one competition

2) 1v1 duel game: super quick access, send a link to anyone in discord and within 10seconds you are in the game. Play wherever you want without download. f2p. no pay to win, just decide who is the truely better player with no excuses

Personally It's a very fun project, I understand a lot of people have a negative opinion about AI being used in game creation. And I get it, I see a lot of "one shot" this one shot that, so many AI Slop being created, I don't really understand what is attractive at doing these one shots.
But in my case, it unlocks something I would have otherwise never been able to achieve. I spent ... maximum 1 month on it so far. Working on it during the evenings when I have some time, and I am able of achieving so much it's insane.
AI is a great instrument that in my case helps apply my creative side instead of taking over the creativity from me. The technical detailed part is where it takes the lead, but I am always the decision maker and I am always in control.

Some examples...
I build the bot on the foundation of the quake 3 bot, based on phd research from van Waveren (rip) https://www.researchgate.net/publication/240430519_The_Quake_III_Arena_Bot
Then I expanded that and split the bot into 3 parts:
- Motoric skills: Aim & movement.
- Atlas: Map knowledge and flight paths
- Brains: Decision architecture and scenario 'reading'/understanding

Set up the whole netcode with the same idea as John C Carmack, set up a replay system, achievement roadmap, and many more.

In the end even if the game is not succesful or people can't get over the fact I used AI - I still learned a lot and most importantly am having a lot of fun...

Anyway sorry for the long writeup this is turning into a blogpost now...

https://scoville.gg/

Looking for playtest feedback!

Still very early stages, but you can play vs the bot on 'Arcadium' and 'The Pit'
Other arenas bot's map knowledge is not mapped yet, will finalize these maps first (geometry wise) and then ready the bot.
Next arena I'm working on in Shoreline (II) -> very heavy load at the moment bear with me, not optimized yet, but trying to make a "nice" looking map. Takes lots of time to get the right assets and textures in place but trying a new workflow.
And of course... I placed the SPIRE arena. Geometric shapes and slopes should feel OK. Even improved the game physics engine moving from box-shapes only to non-box shape that was needed to make the map work. https://imgur.com/a/UBTLmU3
Skybox just a random filler for now, trying to figure out how to make that look good

p.s. yes multiplayer is already live but please be aware the server is in EU. You get around 100ms from east coast USA, just tested. More servers will be placed globally later

Feedback needed:
- Fight the bot on the PIT: general feedback about his movement, rockets, prediction, etc?
- Fight the bot on Arcadium; Lots of improvements to be done on his rocket jump skill and speed. Any observations what makes him easy to beat? What to improve?
- Overall physics & settings feeling: Does it feel very off to tf2? I feel like there is a mouse accell going on that I'm getting changed. Rocket jumps feel slightly weaker than in tf2 - and wall jumps definitely feel less effective. Any input is helpfull here...

Thanks for the interest

Why?

lol me having fun with my virtual company
i can write a comedy sketch about my journey:

https://i.imgur.com/DO8qIpf.png