Check out my persistent decal system that I made last week! I've done one before but I lost the code, so this is from scratch. I also fixed a problem the old one had where decals over the edges of the mesh had gaps in it
1
27
I was baking out the mesh positions to their lightmap UVs. To apply the decal I blit from the whole position map to an overlay map with a custom shader that applied the decal to all the position map areas that it covered
1
27
Added feedback when you finish a route you've already connected. A lot of people kept doing the same route over and over, not realising they weren't making progress! #gamedev #IndieDevs
6
88
Ant Diggle retweeted
Now I've taken a break, I'd like to announce that I am available for consulting.
If you are looking for a proc gen expert to advise on approaches and design, DM me.
I'm also available to give customised support for my tools and libraries, or add features you need.
Been jamming this weekend with @ahmunnaeetchoo, you can tell he's a head animation programmer
I'm using action validity to alleviate some complexity (e.g. you can only jump when you're touching the floor) but you recommended combining actions and letting the Animo figure out the applications right?
2
30
Ant Diggle retweeted
The full recording of their GDC talk "Little Learning Machines Postmortem: A Game About Training Neural Networks" is now live on @Official_GDC's YouTube!
Catch all the behind-the-scenes insights on the making of the game🎮🤖
Watch here: youtu.be/o_NGucEip08?si=9TqJ…
Excellent video about a game with similar motivations to my Neurosmith. I met Dante last year and I'm glad to see both he and the rest of the team have smashed it!
NEW VIDEO
In my latest (sponsored) episode, I sit down with developers from @transformsai to talk about 'Little Learning Machines' (@LoveFearLearn) - a puzzle game where you train AI NPCs called 'Animos' to solve them for you!
Watch on YouTube Now:
youtu.be/keCc1QB7Fvk
1
83
I've finished setting up optimisation for Neurosmith neural nets! It only experiments with the learning rate at the moment but the framework is there to support anything else
1
57
I'm planning to add exploration rate and other common hyperparameters, as well as the NN layer parameters. Really want to make at least the first brain train in under 30s so it's responsive like other optimiser games e.g. Poly Bridge, Opus Magnum
1
43
Submitted our eerie museum game to #gmtkjam with @JamesALockett @btkalman @Grifflon916, really proud of this one! Check it out! fuseinabowl.itch.io/gmtk-24
Ant Diggle retweeted
I'm proud to announce the first ever
AI and Games Conference!
A new event built to support the industry, with two tracks of talks on AI for gameplay and production.
A stack of talks are announced now, with more coming soon. Early bird tickets on sale now! #AIandGamesConf24
Announcing the AI and Games Conference 2024.
Bringing together the leading experts in AI for the video games industry.
- November 8th 2024, hosted @GoldsmithsUoL
- Two tracks of talks on AI in game development.
Early bird tickets on sale now at:
aiandgamesconference.com
I've just discovered that while training my AI rookies would *only* consider invalid actions when thinking about how well they're doing e.g. jumping while they're already in the air... oops! Fixing that has cut training time by 90%
3
117
I've just set up our repo for the #gmtkjam, looking forward to it! I've already set up Github actions to build and deploy it to Itch. One less thing to do during the jam!
1
84








