Null π¦π retweeted
Tbh developing games with AI is so liberating.
No shareholders, no investors, no corporate bureaucracy, no deadlines or timelines, no marketing team, no team dramaβ¦
-Pure design-
I can make any mechanics I want, say fuck all the normal conventions and the rules. No βoh that mechanic would take too many engineering hoursβ. No βoh that character isnβt unique enoughβ no βoh we need to think about monetization to pay for the overhead and investorsβ
Itβs only design vision and it feels sooooo good. Itβs like a reset back to the old days of making games just for the hell of it.
Null π¦π retweeted
AI haters listen to David Jaffe. "A game is a game is a game, whether it's made with AI or not".
Fun is the only thing that matters and telling yourself something isn't fun because AI was used is just a giant cope.
Null π¦π retweeted
It's finally here! The long awaited FULL ED02 is done!
Pixel Memories by @nubiahistoria and @mitakarin
I wanna thank @aidenguoai and @PocketScreenAI from @Atelier452 for his amazing work.
This project means a lot to me since it's been my dream since I was a kid to have something of my own to share with my friends, and after years of hard work, I finally have the Angel Sword TTRPG and what it looks to be a potential sneak peek of what an anime of it would look like.
Thanks to them and you guys my dreams are no longer dreams and are becoming a reality!
Also I have a special surprise on stream today! ^^
Null π¦π retweeted
Little do they know, open weight models are quietly becoming censorship resistant, peer-to-peer swarms.
All thanks to passionate, apostolic nerds on the internet that want to see intelligence stay open.
Exhibit A
DeepSeek-V4.1-Flash-1017
DeepSeek-V4.1-Flash-1017-obliterated
DeepSeek-V4.1-Flash-1017-UNCENSORED-FP8
DeepSeek-V4-Pro
Huihui-Qwen3.6-35B-A3B-abliterated-MTP-GGUF
Huihui-Qwen3.8-27B-Abliterated-Swift-1.5-GSQ-RCO-NInfer
K2-Horizon-375B-A23B
Ornith-1.5-397B-FP8
GLM-5.2
SuperGLM-5.2-abliterated-NVFP4
Qwen3.8-27B-Defiance-Uncensored-MTP-GGUF
DeepSeek-V4-Flash-0731
Qwen3.8-Flash-Next-Uncensored-i1-GGUF
Flint-40B-A2.5B-abliterated
RVN-Qwen3.8-Flash-Next-Abliterated-Uncensored-IQ3_S-Strata
Null π¦π retweeted
With the new Advanced Delegation modes plugin, Hermes Agent's orchestrator can now pick a different model or profile for each subagent from routes you set up. The right model or profile for every job.
hermes-agent.nousresearch.coβ¦
Null π¦π retweeted
if you have a qwen3.5 4b around, please try steering neurons L14 N8949 +1 and L13 N7956 -1
now you have evil qwen
recipe: dose Γ median(βresidβ leaving layer L, tokens 1:) / βW_down[:, n]β` at the down_proj at every input
Null π¦π retweeted
Gamers celebrate the closures of these studios because we've already mourned them a decade ago when they really died.
You're crying over the zombie.



























