AniSora: Open-source anime video generation model
komiko.app
Coding without a laptop: Two weeks with AR glasses and Linux on Android
holdtherobot.com
Show HN: Chat with 19 years of HN
app.camelai.com
Show HN: Turn any workflow diagram into compilable, running and stateful code
workflows.diagrid.io
ARMv9 Architecture Helps Lift Arm to New Financial Heights
nextplatform.com
Experts have it easy (2024)
boydkane.com
Dead Stars Don’t Radiate
johncarlosbaez.wordpress.com
How to have the browser pick a contrasting color in CSS
webkit.org
Bike-mounted sensor could boost the mapping of safe cycling routes
newatlas.com
Every programming language has its 'killer' domain
huijzer.xyz
FreeBASIC is a free/open source BASIC compiler for Windows DOS and Linux
freebasic.net
Confessions about my smart home
frenck.dev
The Lost Japanese ROM of the Macintosh Plus
journaldulapin.com
Push Ifs Up and Fors Down
matklad.github.io
If nothing is curated, how do we find things
tadaima.bearblog.dev
Understanding Transformers via N-gram Statistics
arxiv.org
“Streaming vs. Batch” Is a Wrong Dichotomy, and I Think It's Confusing
morling.dev
Palette lighting tricks on the Nintendo 64
30fps.net
Mice grow bigger brains when given this stretch of human DNA
nature.com
Espanso – Cross-Platform Text Expander Written in Rust
github.com
Memetics – A Growth Industry in US Military Operations (2006) [pdf]
apps.dtic.mil
O2 VoLTE: locating any customer with a phone call
mastdatabase.co.uk
Need to read this asap, but 6.5% is an exceedingly small effect size and not robust given many confounding developmental and technical factors. The typical coefficient of error between volumes of even large brain regions in genetically identical mice is usually well over 5%. And I suspect the transgene inly tested on on genetic background—-usually C57BL/6J.
Will emend this post tomorrow with corrections after reading carefully. I can say with reasonsble assurance that no one has measured and weighed more mouse brains than I have ;-) (except you John Wong)