amanharshx

8 ms·

1.
by amanharshx 2mo ago
Every time they deploy a new model, it feels like older models get slower or dumber. Maybe because of resource allocations, maybe because they started to use quantized models, cant be sure of that. What i found is that you have to adapt you
2.
by amanharshx 2mo ago
The design and layout made it harder to read than it needed to be. Regardless, the inference costs dropping almost 50× is really amazing to see. And now Kimi K3 release has shown how open models are getting closer to the frontier level alre
3.
by amanharshx 2mo ago
Its always the feature combinations that get can get to you. Individually i feel like they make sense, but together they can create some surprising vulnerabilities.
4.
by amanharshx 2mo ago
This is really neat. They've done some really impressive engineering here ngl with the ~95% KV cache hit rates. MiMo and Deepseek both do seem to get the job done for me. Hope they can keep this pace while staying open source.
5.
by amanharshx 2mo ago
WOW! Cant wait to tell to future generations that we had voluntarily made these algorithms to manipulate and influence our own brains
6.
Show HN: Control YOLO Training and Datasets from Claude/Cursor via MCP(github.com)
3 points by amanharshx 2mo ago|0 comments
7.
by amanharshx 2mo ago
You might want to take a look at AionUI. Its a desktop app. I was an early contributor there and they do seem to work on like AI collaborations conversations rather than a single user only chat.
8.
by amanharshx 3mo ago
antigravity is one of the worst tools i have ever encountered with
9.
Show HN: Export YOLO/RF-DETR Models to ONNX, TensorRT, CoreML – 100% Local(github.com)
5 points by amanharshx 3mo ago|0 comments