Articles by joennlae
12

Anthropic disclosed an $80.7B net loss for the first half of 2026 (wsj.com)

12 hours ago joennlae wsj.com
1

Different Sampling Strategies for LLMs (huggingface.co)

2 years ago joennlae huggingface.co
60

Show HN: Stella Nera – Maddness Hardware Accelerator (github.com/joennlae)

2 years ago joennlae github.com
1

Show HN: Maddness – Energy-efficient DNNs without multiplying (arxiv.org)

2 years ago joennlae arxiv.org
31

Show HN: less than 650 LOC trainable GPT only using NumPy (no framework) (github.com/joennlae)

2 years ago joennlae github.com
1

Show HN: Tensorli – NumPy Only Transformer Training (github.com/joennlae)

2 years ago joennlae github.com
3

GithubNext: AI-Powered Auto-Fix Pull Requests for Issues (githubnext.com)

2 years ago joennlae githubnext.com
1

Show HN: Tensorli - GPT-Like Transformers in Pure NumPy (github.com/joennlae)

2 years ago joennlae github.com
2

Nvidia uses internal LLM to support chip design efforts (nvidia.com)

2 years ago joennlae nvidia.com