Just open-sourced POK-Agent 🤖
A Windows AI agent that learns your tasks: a big model plans (System 2), @typesafeai's Jev acts on screen as System 1, and repeated tasks become muscle memory.
Replays: 28% fewer planner calls.
https://t.co/m1CZGDpdlF
#System1#JEV#AIAgents
@DeepSeekHarness Love the word "harness." Ours harnesses the actual computer sees your screen, drives mouse & keyboard, and repeated tasks become muscle memory so the big model plans less over time. Local-first, Rust+Tauri, open source 🦾 https://t.co/m1CZGDpdlF
#deepseek#agent@deepseek_ai
@dhruvalgolakiya Just open-sourced POK-Agent 🤖 A Windows AI agent that learns your tasks: a big model plans (System 2),
@typesafeai
's Jev acts on screen as System 1, and repeated tasks become muscle memory. Replays: 28% fewer planner calls. https://t.co/m1CZGDpdlF #System1#JEV#AIAgents
@tinyhumansai@NousResearch Just open-sourced POK-Agent 🤖 A Windows AI agent that learns your tasks: a big model plans (System 2),
@typesafeai
's Jev acts on screen as System 1, and repeated tasks become muscle memory. Replays: 28% fewer planner calls. https://t.co/m1CZGDpdlF #System1#JEV#AIAgents
@sundarpichai Just open-sourced POK-Agent 🤖 A Windows AI agent that learns your tasks: a big model plans (System 2),
@typesafeai
's Jev acts on screen as System 1, and repeated tasks become muscle memory. Replays: 28% fewer planner calls. https://t.co/m1CZGDpdlF #System1#JEV#AIAgents
Also in v0.1: offline voice input, image/file drag and drop, and hard safety limits (no passwords, UAC, or calls you didn't ask for).
Rust + Tauri, Apache-2.0. Windows installer 👇
https://t.co/QuqtJiKITh
#OpenSource#ComputerUse
Just open-sourced POK-Agent 🤖
A Windows AI agent that learns your tasks: a big model plans (System 2), @typesafeai's Jev acts on screen as System 1, and repeated tasks become muscle memory.
Replays: 28% fewer planner calls.
https://t.co/m1CZGDpdlF
#System1#JEV#AIAgents
Built local-first for @NVIDIAAI RTX GPUs: run the planner in @lmstudio or @ollama and a local System 1 (Laya) on CUDA, so your screen never leaves your PC.
Next: fine-tuning Laya on our own runs. GPU help would speed this up a lot 🙏
#LocalLLM#NVIDIA