Hi, machine learning specialists!
Today we’re stepping a little outside while True: learn() to share the biggest news from the project our team is actively working on right now.
It’s called SuperWEIRD. It has a different world, clumsy robots, production chains, and considerably more suspicious goo, but the basic principle may feel familiar: you build a clever system, press the button, and immediately discover several outcomes that were definitely not in the training dataset.
https://store.steampowered.com/app/3818770/SuperWEIRD/
For anyone seeing it for the first time, SuperWEIRD is an automation roguelite about building production chains, commanding robots, rebuilding a broken planet with Tetris-like pieces, and trying to stop the Corruption known as Goofy Goo.
SuperWEIRD now has co-op
We’ve just launched our very first proper co-op build.
Naturally, everything started falling apart almost immediately.
Shortly before release, we discovered that a player could invite themselves into their own co-op session. Which is technically multiplayer, depending on how philosophical you feel about it.
We fixed that, fixed quite a few other things, and probably generated several exciting new edge cases for future us to investigate. Now we need to see how co-op behaves outside the carefully controlled environment of our development machines.
We’re especially interested in full video playthroughs. Written feedback tells us what went wrong, but recordings help us see where players hesitate, what feels unclear, what unexpectedly works, and which robot has decided to produce results with suspiciously low validation accuracy.
You can sign up for the co-op playtest here:
Players who can record their playthrough will be invited first.
And one more very important thing
Our artist Anya has brought one of the game’s NPCs to life — a character who has already quietly become the project’s little mascot.
Please take a moment to admire this adorable creature and send some well-deserved compliments Anya’s way.
We’ve also uploaded the painting process to the SuperWEIRD YouTube channel, for anyone curious to see how it all came together.
while True: learn() taught us that even the smartest system behaves differently once real people start experimenting with it. That lesson is proving extremely useful again — now with co-op, robots, planetary reconstruction, and Goofy Goo added to the dataset.
Want even more automation?
- Sign up for the closed playtest of our new game
- Join our Discord
- Visit our website
- Play our free games
- Write a comment under your favorite gaming YouTuber’s video asking them to review the game
Want more automation?
- Sign up for the closed playtest of our new game
- Join our Discord
- Our website
- Our free games
- Write a comment under your favorite gaming YouTuber’s video asking them to review your game