tenpa0000

4 ms·

1.
Show HN: 3-path agent memory – 8 KB recurrent state vs. 156 MB KV at 10K tokens(github.com)
1 points by tenpa0000 7mo ago|0 comments
2.
by tenpa0000 7mo ago
I run Llama 3.2 3B locally for latency-sensitive classification (sub-50ms, so no room for bigger models). At that scale Q2_K vs Q4_K_M isn't just smaller — Q2 starts flipping yes/no answers that Q4 gets right. Not often, but enoug
3.
by tenpa0000 7mo ago
This is great timing — I've been putting off making animated diagrams for a blog post because the Manim setup was too much friction for what I needed. Tried the live demo and the 3D orbit scene is surprisingly smooth. Curious about a c
4.
by tenpa0000 7mo ago
The deduplication problem is the part I haven't worked out cleanly. The hierarchy in veronica-core sidesteps it as long as you declare parent-child relationships upfront — B's spend rolls directly into A's ceiling without a s
5.
by tenpa0000 7mo ago
The flip side is enumerable IDs. Back when I was scraping a site for a side project, sequential photo IDs were basically a free sitemap. YouTube's random-ish IDs aren't just branding — they at least make bulk harvesting annoying.
6.
by tenpa0000 7mo ago
.git/hooks is underrated. I have a pre-push hook that runs my test suite — annoying to set up the first time but I've probably avoided a dozen broken CI runs by now.
7.
Wrapping Code Comments(matklad.github.io)
2 points by tenpa0000 7mo ago|0 comments
8.
Linux 7.0 Makes Preparations For Rust 1.95(phoronix.com)
6 points by tenpa0000 7mo ago|0 comments