PLATE BLOGBlog
Notes
Mostly about running models locally, and what we learned building an agent IDE around hardware people already own.
- 4 min read
What a 4 GB GPU can actually do
Most of this was built on a GTX 1650. Here is what that hardware can and cannot do, without the marketing gloss.
- 5 min read
Which local coding model should you run?
A straight answer organised by how much VRAM you have, and what actually changes as you go up.
- 4 min read
Why the whole IDE is free on local models
Every AI editor charges per seat because every keystroke costs them inference. Move the inference to your machine and that whole argument disappears.