a chatbot not a coding agent.
I have another RSI~ omnimodal coding agent I will share next. for natural bot 2 bot collaboration between the two... or more.
talk to yr bro, get them to deligate in chat to the coding enabled bro.
My defaultmode inspired memory module for agents has been a bit of a sleeper... but because RSI is so antagonistic to claude, I figured I'd finally implement nightly fine tuning of the model that runs it.
Using DPO styled training per night on the memindex, and evictions.... π§ π΄
running last of the needle evals and I'll update the repo...
next push comes with well documented hooks. cli, tui, webgui and discord as current platforms for multimodal i/o.
local / cloud model provider support.
smol bunch of tools for todo planner, bookshelf and yt/doc parsing.
I had my FOF emulation on aderpy for a while with a todo for adding a vocaloid style midi editor as a WIP...
You can now import midi, for melodyne style voice synthesis editing and export...
is SAMchan a thing?
In the 2010s, NZ producer Jeremy Coubrough, aka Tlaotlon, made some very interesting abstract techno. Ten years on, I revisited his discography for Mixmag ANZ.
https://t.co/6UDr7PlpWX
https://t.co/3UdDxlJsGm
I brought this album in 1999, based on a hunch. It is my fav album OF ALL TIME, and everytime I remember that and I fall into it I feel so at home in space and time.
Introducing π»ππππππ²ππππ, our ~74M-parameter streaming audio codec powering LFM3-Audio-2B and LFM3-Omni-2B.
An audio codec turns speech into tokens a language model can predict, then converts those tokens back into audio. We built ours around two requirements: tokens that are easy to predict and fast conversion back into speech.
On a single CPU thread on an Apple M4 Max, LiquidCodec runs at approximately:
> 361Γ real time for encoding
> 87Γ real time for decoding
Both measurements use FP32 precision.
1/ π§΅
todays haul...
i also brought a sata/ide pci adapter for my pentium build...
about to read this sue bradford history of UBI policies in nz πββοΈ