
software, gamedev, ai | making games @ankkalab
not sf
Joined March 2010
- Tweets31.1K
- Following7.2K
- Followers7.3K
- Likes189K
Interesting new model, supposedly beats Qwen 27B
🚀Excited to announce Nanbeige4.3-33B!
To push the performance ceiling of small agents, our strategy in Nanbeige4.3 starts with building a powerful teacher first.
Nanbeige4.3-33B delivers SOTA agentic performance among open models at ~30B parameters:
* Strong results on agentic coding and cowork benchmarks.
* When integrated into mainstream agent harnesses, it tackles real-world tasks with cost-efficient inference.
* It will serve as a teacher for upcoming smaller Nanbeige models, starting with Nanbeige4.3-3B.
huggingface.co/Nanbeige/Nanb…
huggingface.co/Nanbeige/Nanb…
#LLM #Agent
CPU cooler now shows if my 4090 is running an AI workload or a game. The RGB color scheme changes accordingly.
I can use the PC as a remote AI server and kids won't be able to start games on it while it's doing work
Since GPT-3, OpenAI have argued they can't release model weights because it would be too dangerous or whatever. Today there is an open model from China that beats their current flagship.
Absolute humiliation.
you're so real for that. but there is another, more subtle evil lurking which I must warn you about.
platforms curated by algorithms may push humans to write algorithmically optimal content.
if you let it control you, you may see yourself turn into a bot without ever using one
Discord moderation I can trust. Built-in AutoMod flags certain words, then Opus 5.5 uses the context of the conversation to make a decision based on the user's intent.
In this case, no action was taken.
making an account to a website with email and password, it should first try logging in to that email using that password, and reject if it succeeds
Vibe your way to 50K.
The 2026 Epic AI MegaJam begins Nov 12: epic.gm/epic-ai-megajam-2026















