My way in contributing to Superintelligence advocates is kvcachestore.com/ by pre-computing KV cache on files, you can save a lot on computational costs with data you repeatedly send through an LLM. It uses a cli and has an optional online storage service you can use.
Sometime soon someone is going to be able to fan out NVIDIA Workstations as L1-L3 cache clusters and it is going to be amazing historically unprecedented compute. For example a 10-node, 80-GPU Vera Rubin cluster with 23 TB HBM4 and 800G InfiniBand scale-out can handle up to 36T trillion parameters in NVFP4 models and train 30B–70B dense foundation models from scratch and it can only get optimized from there. I would never watch TV again if I had this.
@llm_wizard
We release Whistle: speech to text in a 16.9MB file, running on the CPU in the same engine as Needle. It mostly beats Whisper base with 9x less file size and 6x speed. Whistle supports English, German, French, Spanish, Italian, Dutch and Polish.
cactuscompute.com/blog/whist…
And here's Jev playing Tetris
(Btw I am not trying to hype Jev, I am just saying that there are lots of quick Jev clones, but they don't capture what makes Jev so popular/useful).
I honestly don’t care at all that I spend 30 years of my life learning syntax and now that knowledge is essentially useless.
Frankly, I’m relieved that I don’t need to care about it anymore. That stuff never really mattered anyways. Code was always just a means to an end.
All I care about is building products that users love. Now I can do that faster. That means I can have bigger impact.
If anything, software engineers should all be celebrating. I don’t understand why you’d “mourn” not needing to write code. The code was never the point.
Electronics & wiring are the trickiest part of assembling a robot. We’re shipping an entirely new experience for electronics in your @tnkrdotai repository.
Goal is to make it 10x easier to document your electronics system & 10x easier for anyone to assemble.
Every team / builder needs this.
today at @nianticspatial we're launching the Places Library: 100 real environments to train robots in, with even more coming soon.
warehouse, last mile, residential, and more. every one a simulation-ready USDZ. start here, then capture the sites you actually deploy into.
here's a complexity test to run for your favoorite LLM: ask it to write its own alternative to wayland or x11 and see what it comes up with for a window manager server.
I made a rust screensaver for the terminal. its highly customizable including an experimental llm integration to make custom screensavers for whatever you want github.com/awdemos/phosphor