Show HN: I trained a 125M model to autocomplete piano on-device
(news.ycombinator.com)
1.
2.
Microgpt in pure C hits 10M tps on Apple m5
(news.ycombinator.com)
3.
Qwen 3.8 27B
(news.ycombinator.com)
4.
The Download: the next big thing in LLMs and how AI academic research is shifting
(technologyreview.com)
5.
These startups are chasing the next big thing in LLMs
(technologyreview.com)
6.
An SLM trained on $8 ESP32-S3
(news.ycombinator.com)
7.
Our favorite memories at the movies
(theverge.com)
8.
TokenTown: A visual way to understand how LLMs work
(news.ycombinator.com)
9.
Transformer Transformer: A Unified Model for Motion-Conditioned Robot Co-Design
(news.ycombinator.com)
10.
Running Kimi K3 on a M1 Max
(news.ycombinator.com)
11.
Running Kimi K3 on a M1 Mac
(news.ycombinator.com)
12.
13.
14.
The Must-Have Exclusives From San Diego Comic-Con 2026
(gizmodo.com)
15.
16.
14× faster embeddings: how we rebuilt the ONNX path in Manticore
(news.ycombinator.com)
17.
Is One Layer Enough? A Single Transformer Layer Matches Full-Parameter RL Train
(news.ycombinator.com)
20.
Generating Images with a 2025 Android
(news.ycombinator.com)
21.
22.
23.
24.
DiffusionBench: Towards Holistic Evaluation of Generative Diffusion Transformers
(news.ycombinator.com)
25.
Krea 2: SOTA open-weights 12B image model
(news.ycombinator.com)
26.
Krea 2 Technical Report
(news.ycombinator.com)
27.
Overfitted a 900KB Transformer to Compress a 100MB CSV into 7MB
(news.ycombinator.com)
28.
Unlimited OCR: One-shot long-horizon parsing
(news.ycombinator.com)
29.
Keanu Reeves Is Helping to Build a New Lego Movie
(gizmodo.com)
30.
LLMs Are Complicated Now
(news.ycombinator.com)
Today's top topics:
blade runner
ridley scott
neo-noir
fan art
writing
authors
fourth wing
nvidia
vernor vinge