I feel that the trend towards training models to autonomously go off and try to do everything themselves is anti-human. We should, IMO, be training LLMs to support humans in their learning, creativity, and iterative experimentation.
GENERATIVE AI
-
Inquiry about safetensors version availability vs ggufs
By
–
Do you have a safetensors version published or only ggufs?
-
Anthropic engineers’ token-saving habits, no setting changes needed
By
–
this guy literally breaks down the exact habits Anthropic engineers use to save millions of tokens without changing a single setting 🤯
— Charly Wargnier (@DataChaz) 22 mai 2026
Watch the video, then bookmark the written guide 👇 https://t.co/gAHjGmFxwy pic.twitter.com/Hlb96srixBthis guy literally breaks down the exact habits Anthropic engineers use to save millions of tokens without changing a single setting Watch the video, then bookmark the written guide
-
GPT-5.5 Generates 30k Lines QML with Agentic Reasoning
By
–
GPT-5.5 cranking out 30k lines of QML for the Omarchy 4 branch + nailing subtle agentic reasoning!!
-
Claude MD File: Top GitHub Repo Uses Karpathy’s 4 LLM Principles for AI Control
By
–
DID YOU KNOW THE #1 GITHUB TRENDING REPO (146K+ STARS) IS LITERALLY JUST A CLAUDE MD FILE? It uses @karpathy
's 4 LLM principles to keep AI in check: → Seek clarity: Always ask before making assumptions.
→ Stay minimal: Avoid bloat and keep things simple.
→ Edit -
Thousands of tokens per second across parallel requests
By
–
1000s of toks/sec across a dozen parallel requests if not more
-
Sarcastic commentary on short prompt and costly hardware
By
–
I mean, dude's longest prompt was 368 tokens and longest output was 4k He's literally crying next to that hardware and hoping to get enough engagement that elonbux can pay for his next month installment on those desk warmers
-
On-chain GANs: InChainPepeGAN, Concrete, and AAA improvements
By
–
I stand corrected! There are at least two on-chain GANs that predate this, InChainPepeGAN (from 2023!) and Concrete by Higgs by @0xdiid
. So AAA is not first in that regard, although it's still innovative, as it generates clear subjects in high res, so it pushes fidelity forward -
GPT 5.5 improving while Claude models worsening, no clear winner
By
–
GPT 5.5 seems to be improving in that direction now, and Claude models are getting worse at it, so I don't think there's a clear winner now.
-
Need better evaluation of models for human-AI cooperation
By
–
We desperately need better ways of evaluating models. Something that shows how helpful they are at working hand-in-hand with humans to help them get stuff done in a cooperative/iterative way. The Claude models have consistently been better at this, and the market rewards that.