LLMs Are Complicated Now
(news.ycombinator.com)
1.
2.
IEEE Rolls Out Large Language Models Virtual Training Course
(spectrum.ieee.org)
3.
Integer Quantization: Deep Dive
(news.ycombinator.com)
4.
GateGPT: 56k tokens per second Transformer (KV cache) on FPGA at 80 MHz
(news.ycombinator.com)
5.
Deficient executive control in transformer attention
(news.ycombinator.com)
6.
Forget FAANG—there’s a new powerhouse acronym for tech stocks in the AI era: MANGO
(feeds.feedburner.com)
7.
Vacuum-Form Signage
(news.ycombinator.com)
8.
Scientists Find Chunk of Lost Planet in Desert
(futurism.com)
9.
'5D' platformer Screenbound arrives on September 10
(engadget.com)
10.
Transformers are inherently succinct
(news.ycombinator.com)
11.
12.
13.
1-Bit Bonsai Image 4B Image Generation for Local Devices
(news.ycombinator.com)
14.
CODA: Rewriting Transformer Blocks as GEMM-Epilogue Programs
(news.ycombinator.com)
15.
Show HN: KVBoost – chunk-level KV cache reuse for HuggingFace, 5–48x faster TTFT
(news.ycombinator.com)
16.
Show HN: CPU-only transcription for YouTube, TikTok, X, Instagram videos
(news.ycombinator.com)
17.
18.
Nintendo keeps finding new ways to reinvent platformers
(theverge.com)
19.
When Fast Fourier Transform Meets Transformer for Image Restoration (2024)
(news.ycombinator.com)
20.
When Fast Fourier Transform Meets Transformer for Image Restoration
(news.ycombinator.com)
21.
GenCAD
(news.ycombinator.com)
22.
Autoregressive next token prediction and KV Cache in transformers
(news.ycombinator.com)
23.
The shock of seeing your body used in deepfake porn
(technologyreview.com)
24.
After 8 years, I rewrote my open-source PyTorch curvature library
(news.ycombinator.com)
25.
26.
27.
28.
A polynomial autoencoder beats PCA on transformer embeddings
(news.ycombinator.com)
29.
Train Your Own LLM from Scratch
(news.ycombinator.com)
30.
Transformers Are Inherently Succinct (2025)
(news.ycombinator.com)