High-res image reconstruction, latent diffusion models from human brain activity(biorxiv.org)

88 points by rntn 3 years ago | 17 comments

CoQ4 is a functional substitute for CoQ10 and can be targeted to mitochondria(biorxiv.org)

79 points by birriel 3 years ago | 39 comments

Widespread drastic reduction of brain myelin upon prolonged endurance exercise(biorxiv.org)

78 points by anigbrowl 3 years ago | 34 comments

Sitting for a long time shrinks your brain even if you exercise(alz-journals.onlinelibrary.wiley.com)

60 points by codexon 1 year ago | 25 comments

QLoRA: Efficient Finetuning of Quantized LLMs(arxiv.org)

270 points by Garcia98 3 years ago | 90 comments

Why flying insects gather at artificial light(biorxiv.org)

256 points by Luc 3 years ago | 104 comments

Pattern matching beats narrative reasoning(blog.scottstevenson.net)

36 points by tosh 9 days ago | 20 comments

Scott Stevenson's Substack essay arguing that good bets come from pattern matching rather than narrative reasoning and can't be easily explained in words.

The problem is not AI code, but not knowing about system architecture or intent(ssp.sh)

317 points by zazuke 28 days ago | 204 comments

A short note in a personal knowledge base arguing that the real problem with AI-generated code is that teams no longer understand their system architecture or the intent behind design decisions.

Quality non-fiction books are the antithesis of AI slop(resobscura.substack.com)

463 points by benbreen 3 months ago | 222 comments

Benjamin Breen's Res Obscura essay arguing quality non-fiction books counter AI slop, introducing a vibe-coded tool indexing prize-winning books.

Some critical issues with the SWE-bench dataset(arxiv.org)

331 points by joshwa 1 year ago | 113 comments

arXiv paper introducing SWE-Bench+, an empirical analysis revealing data quality issues in the SWE-bench coding benchmark for LLMs, such as solution leakage and weak tests.

Prevent cognitive debt by manually retyping LLM-generated code(ankursethi.com)

502 points by mpweiher 2 months ago | 413 comments

A personal blog post arguing that developers should manually retype code generated by LLM coding assistants to keep understanding it and avoid cognitive debt.

IRS tests free e-filing system that could compete with tax prep giants(washingtonpost.com)

693 points by elicash 3 years ago | 349 comments

Washington Post article reporting that the IRS is testing a prototype free government e-filing system that could compete with commercial tax preparation companies.