A rather geeky/technical weblog, est. 2001, by Bramus
To be launched later this year:
LEGO Super Mario is ready to run and jump in a new real-life game. Create LEGO brick challenges for you and your friends, defeat enemies and of course, collect coins!
Pretty sure my son will like it as much as I am.
Press Release: The LEGO Group and Nintendo partner to take legendary brick-building to a new level →
During the winter break Kevin Hughes decided to try and train an artificial neural network to play MarioKart 64:
After playing way too much MarioKart and writing an emulator plugin in C, I managed to get some decent results. Getting to this point wasn’t easy and I’d like to share my process and what I learned along the way.
TensorKart: self-driving MarioKart with TensorFlow →
TensorKart Source (GitHub) →
Because you’re not nagged with screens like this:
Nor screens like this:
I’ve been playing Super Mario Run the past few days and haven’t regretted paying $9.99 to get the full version. Yes, that’s above average when compared to other apps, but at least you get a real game in return for it:
More photos: What Super Mario Run would look like as a free to play game →
(*) If you just jump around a bit you can speedrun-finish the game in one single day. However, one week in and I’m not even halfway because collecting all special coins will take quite some skills (and time).
I’m excited about this one. Still one week to go before its release
A new Mario game you can play with one hand! On December 15 of 2016, Mario jumps into a new adventure on iPhone and iPad.
Extract the data from the outdoor levels of Super Mario Bros, represent it as text (as shown above), feed it to a neural network, let it iterate the data quite a few times and *BOOM*, you get this:
Remember that Super Mario World “Credits Warp”? The guy’s back, and this time he has injected the source code of Flappy Bird into Super Mario World by just playing the game:
Using various Super Mario World glitches, I injected the code for Flappy Bird
MarI/O is a program made of neural networks and genetic algorithms that kicks butt at Super Mario World
The source code is freely available.
Fascinating trick in which one actually reprograms Super Mario World by just playing it to make it start playing the end credits upon triggering a crash:
The goal of this run is to trick the game into playing the credits in the first level. It works by using a glitch to allow yoshi to eat a chuck, which you cannot normally do. This executes improper code and usually crashes the game. The memory addresses read during the crash can be manipulated. By spawning a p-switch in a certain order and placing it at a pixel perfect position, as well as despawning two block break sprites at a mostly pixel-perfect height, the game reads the x/y position of the sprite slots as though they were code instead of executing random code and crashing. The shells at the start of the level are spawned in these sprite slots and placed at pixel perfect positions which, when read as code, call the credits.
A much shorter run, but without any explanation, is also available.
More on this Credits Warp thing:
How does Super Mario Bros teach you to play purely through level design? Game Designer Dan “OtherDan” Emmons breaks down the mechanics and offers a lesson in game design techniques.
Whatever you do: you’ll get the mushroom.
Mari0, the game that cleverly mixes Super Mario Bros. and Portal, has been released over the weekend. Available for Mac, Windows & Linux. Source included (written in LÖVE)
UPDATE The 1.1 download seems broken on OS X. Can be fixed by viewing the package contents and deleting mari0_1.0.love from the Resources Package Folder