AI is moving onto your computer.
With Scribbler, run AI directly in your browser. No server. No install. No login.
With ScribblPad, turn an AI-capable PC into a ready-to-use AI computer with workspaces for coding, design, creation and analysis.
#AI#OnDeviceAI#AIPC#LocalAI
LocalAI 5.0 is out: run open models for chat, images, speech and video on ur own machines, with an API that speaks both OpenAI and Anthropic formats. new desktop app, no GPU needed
it does the work, u keep the files
https://t.co/K0Ty1gDn2K
How is this not a fraud?
India's leading AI lab worth $1.4 billion got funded and GPU to build a sovereign model in India by the Government.
They have now stopped innovating.
..and using the same GPU to sell Chinese model openly.
This is absolutely rubbish and mockery.
every open weights release moves a little more power from the few labs that can train frontier models to the many people who can build with them
that's why the license line matters as much as the benchmark table
Thinking Machines just open-sourced Inkling 2 Small, Apache 2.0 weights
the Small is 594B parameters. naming things really is the hard problem in computer science
https://t.co/6fBXHXKjC2
llama.cpp just hit v1.0.0
the engine that took LLMs from datacenter-only to running on ur laptop, ur phone, a 2015 kindle and a 1GB TV box. 12,000+ builds later, they finally called it 1.0
https://t.co/nO3Z4ygrFW
Many times i like to quickly jot down something (notes/thoughts.diagrams etc).. I don't like installing apps or signing up for some service. so I've built a small "whiteboard" tool.. you can draw with a pen, write some text, and save whiteboards.. all the saved whiteboards will remain in the browser - there is no sever or backend.. so everything is completely private.. and you can also download a whiteboard and share with others who can load in the same app.. do let me know if you find it useful -- https://t.co/VTY5YoMRh7
KORA unveiled a Linux desktop with AI agents baked into the OS, engineered with Qualcomm to run a 26B model entirely on-device on Snapdragon PCs. local by default, cloud by exception
https://t.co/m8IO6OMQ3e
local AI doesn't have to win every benchmark. it has to do a useful job on the machine u already own, at a cost u can live with. that's a different engineering target
UC Berkeley open-sourced FreeToken!
FreeToken is an edge-native inference engine for running frontier-scale MoE models on consumer hardware. GPU, CPU, host memory, and PCIe interconnects treated as one unified inference platform.
Nace AI's Drex Max v1 is a 27B decision model for routing, triage and policy checks, reported to run offline on Apple Silicon. a reminder that useful local AI doesn't always need to write an essay
https://t.co/TWT6i7KdpW
someone built a 1.2m open-source humanoid that runs on a Raspberry Pi 5 and a Radxa CM5. CAD, firmware and the walking policy are all published, and on day 366 it was walking around a public event in Singapore
https://t.co/juRe8pRNHE