771–780 of 913 posts
https://operator.io - multiple isolated agents in Telegram with their own memory and tasks has been great for automating reminders, keeping tabs on things, and acting as a personal exocortex
I recently built a physics-driven auto-panner based on PONG, so I can also play PONG inside my DAW. https://portwaydsp.com I also built https://plugins.audio - a Dribbble-esque showcase of audio plugins. It's been a dream to be able to work in the audio software space. I've been a musician, designer, coder all my life, and had a few moonlights with building audio plugins over the years. But now with AI at hand, I can…
I've been making a lot of audio experiments for my own amusement. They all have some potential to drain your cpu, sorry! https://www.noisetable.xyz/ - a collection of chance-based audio 'channels' in a VCR inspired interface https://concrete95.net/ - a musique concrète web app that's made to look like windows 95. Pulls random audio from freesound.org and loops a small section, you can also layer synth pads or melodic…
I shred guitar in the evenings to focus on technique, but a hectic day job means I get almost no time to twiddle knobs on my amp sim chasing the right tone. So I built a tool where I just prompt "Brian May Solo Bohemian Rhapsody" — the LLM fills a small JSON contract and feeds to a script that generates the XML, and I load it straight into the plugin. Two minutes and I feel like I'm tearing it up at Wembley '85 :)...…
Main components are: - a router script (onpro) that probes the box and runs the command there, falling back to local automatically if it's asleep or unplugged. - a PreToolUse hook in my AI coding setup intercepts the heavy commands like builds, test suites, ffmpeg, OCR, big pandas jobs. Forces them onto the workhorse instead of melting my MacBook Air.
The agent decides what to run, and a deterministic layer guarantees the heavy stuff lands on the right machine.
Testing Workhorse a bit more before sharing.
Also https://motate.app/ was an attempt to make math writing easier for me (as I recently needed to hold my newborn through most of my LinAlg refresher course) but I've since started trying to figure out how to make it useful for broader science/math writing and education.
Earlier quoted context omitted.
Wait so this makes it so I can use my DDR5 as well as my VRAM combined? This is actually sick if so. Maybe I will actually have to go out and buy some more DDR5 (currently only have 32GB...)
Yep, thats it what it does. Only works with nvidia. The difference it does use safetensors, and not gguf's. But it does dynamically requant to int4 8 or bf16.
I shred guitar in the evenings to focus on technique, but a hectic day job means I get almost no time to twiddle knobs on my amp sim chasing the right tone. So I built a tool where I just prompt "Brian May Solo Bohemian Rhapsody" — the LLM fills a small JSON contract and feeds to a script that generates the XML, and I load it straight into the plugin. Two minutes and I feel like I'm tearing it up at Wembley '85 :)...…
Very cool! I was thinking of doing this. One question, what made you choose polychrome dsp over something like neuraldsp?
My favorite is: https://github.com/theo-sardin/instant-switcher Instant windows switcher with custom shortcuts and instant "opt+tab" and trackpad switching. Simple does exactly what I need it to do (just bypasses the slow window switching that is annoying), with no additional features or bloat. I've also done a TUI that combines my messages from WhatsApp + messenger + discord which is pretty handy at work.
Earlier quoted context omitted.
Yep, thats it what it does. Only works with nvidia. The difference it does use safetensors, and not gguf's. But it does dynamically requant to int4 8 or bf16.
Wow that's actually sick as hell, somehow hadn't heard of this. maybe I will go and blow $700 on a new ram kit... thanks for sharing!
But go try it out now with a 35B model on your current hardware.
Right now, I have loaded qwen3.6-35B-A3B, 128k context, kv cache 2.5GB, thinking. Int8
Using 11.5GB gfx ram, 42GB system ram.
I dont want to oversell. All GPU would be faster, but creating a semi-unified system is deffo a game changer for me.