GreenGames

7 ms·

1.
We got DSV4 Flash 0731 to 32.7 tok/s on AMD Strix Halo(lucebox.com)
5 points by GreenGames 1mo ago|0 comments
2.
AMD Lucebox Beats Nvidia DGX Spark by 3.63x on DeepSeek V4 Flash(lucebox.com)
4 points by GreenGames 2mo ago|0 comments
3.
DeepSeek V4 Flash, up to 32 tok/s on AMD Ryzen AI MAX+ 395(lucebox.com)
3 points by GreenGames 2mo ago|1 comments
4.
Laguna XS 2.1 33B on a RTX 3090: 296 tok/s peak, 152 tok/s at 256K context(lucebox.com)
20 points by GreenGames 2mo ago|1 comments
5.
256K context with 72MiB of KV cache on the GPU(github.com)
3 points by GreenGames 3mo ago|0 comments
6.
A 35B MoE on a 16 GB GPU, without the offload tax(lucebox.com)
18 points by GreenGames 3mo ago|0 comments
7.
PFlash: 10x prefill speedup over llama.cpp at 128K on a RTX 3090(github.com)
3 points by GreenGames 5mo ago|1 comments
8.
by GreenGames 5mo ago
This reads like you didn’t read the post. z-lab runs BF16 on B200 (54+ GB). There is no z-lab path that fits on a 24 GB 3090. That is literally the entire point of our work, and it is stated in the second paragraph. If you had checked the H
9.
We got 207 tok/s with Qwen3.5-27B on an RTX 3090(github.com)
165 points by GreenGames 5mo ago|52 comments
10.
by GreenGames 5mo ago
We built a standalone C++/ggml speculative decoder for Qwen3.5-27B Q4_K_M with a DFlash block-diffusion draft. 207.6 tok/s peak (5.46x over AR); HE 10-prompt bench averages 129.5 tok/s at DDTree budget=22, single RTX 3090, 24
11.
Show HN: OS Megakernel that match M5 Max Tok/w at 2x the Throughput on RTX 3090(github.com)
6 points by GreenGames 5mo ago|1 comments
12.
by GreenGames 10mo ago
Super interesting take Paul. Curious btw, how are these teams actually encoding their “institutional knowledge” into constraints? Like is it some manual config or more like natural‑language rules that evolve with the codebase?
13.
Cua (YC X25) is hiring an engineer(ycombinator.com)
1 points by GreenGames 1y ago
14.
App-Use, Control Individual Applications with CUA Agents(trycua.com)
2 points by GreenGames 1y ago|1 comments
15.
by GreenGames 1y ago
Hi there, Alessandro and Francesco here. We just launched an experimental feature in C/ua called App-Use. It lets you create virtual desktops scoped to specific apps (e.g., "Safari and Notes only") to give your agents focused
16.
Show HN: Lumier – Run macOS VMs in a Docker(github.com)
159 points by GreenGames 1y ago|52 comments
17.
by GreenGames 1y ago
Sorry my bad I didn’t see that. Should I cancel it?
18.
Microsoft is reportedly about to lay off 3% of its workforce(techcrunch.com)
13 points by GreenGames 1y ago|2 comments
19.
Polaris is giving free GPUs/CPUs for everyone(polariscloud.ai)
3 points by GreenGames 1y ago|0 comments
20.
Improvements in reasoning AI models may slow down soon, analysis finds(techcrunch.com)
3 points by GreenGames 1y ago|0 comments
21.
by GreenGames 1y ago
I don't understand what Klarna is doing here lol
22.
Microsoft and OpenAI are renegotiating their partnership(techcrunch.com)
3 points by GreenGames 1y ago|0 comments
23.
Apple developing new chips for smart glasses, Macs, and more(techcrunch.com)
6 points by GreenGames 1y ago|0 comments
24.
by GreenGames 1y ago
Sundar just started following cursor on twitter lol
25.
by GreenGames 1y ago
Oh okay thanks - that would be fire tbh
26.
by GreenGames 1y ago
This is very cool! Does your MCP server preserve cookies/localStorage between steps, or would developers need to manually script auth handshakes?
27.
by GreenGames 1y ago
very cool, I know they are all a bit different, but what was the one the surprised you the most and you'd recommend to really try out?
28.
Show HN: Tlume – a CLI tool that converts Tart VM images for Lume(github.com)
5 points by GreenGames 1y ago|2 comments
29.
Show HN: Lume – OS lightweight CLI for MacOS and Linux VMs on Apple Silicon(github.com)
309 points by GreenGames 2y ago|75 comments
30.
Show HN: List of AI Agents(github.com)
4 points by GreenGames 2y ago|1 comments

More ›