.png&w=256&q=75)
🤓 Latest Submissions
.png&w=828&q=75)
RoboProof: Motion-Learning Lab
RoboProof is a physical-AI research prototype for improving how a VEX V5 robot moves. Instead of asking a language model to control motors directly, it combines a Guiding Vector Field (GVF), which provides a structured path-following reference, with a TD3 reinforcement-learning policy that learns residual adjustments from simulated sensor feedback. The system uses the project’s original Simulator/iraLIB so candidate policies can be compared with the existing controller and an untrained policy under repeatable conditions. Our primary learned candidate is a 512-wide, 10-hidden-layer network. We are training it on an AMD Instinct MI300X and preserving run data, checkpoints, and evaluation evidence. The goal is to determine whether learning improves movement reliability and efficiency—not to assume that a larger network or more GPU utilization guarantees better results. RoboProof is still experimental. Training is underway, and the learned policy has not yet passed the independent qualification gate. We are looking for robotics, controls, reinforcement-learning, and simulator contributors to help valid
18 Oct 2026
PCMANai - Self-Learning AMD Hardware Optimizer
PCMANai is an autonomous, self-learning PC optimizer built for AMD hardware. It uses a custom Deep Learning Neural Network that monitors real system telemetry — CPU, GPU utilization, real GPU frametimes via PresentMon's DirectX hooks, RAM, disk, and network ping — to detect performance anomalies and autonomously fix them. The AI selects from 18 hardware-level tools including AMD-specific tweaks: disabling ULPS to prevent GPU throttling, clearing corrupted AMD shader caches, toggling HAGS, disabling MPO for better frame pacing, plus aggressive MMCSS thread priority, Nagle's Algorithm disabling for lower ping, Win32 priority separation, and memory compression control. What makes it truly intelligent is the reward system. After each tweak, the AI measures CPU, RAM, and frametime before vs after, calculating a weighted mathematical reward. Negative rewards punish bad actions. Over time, the Neural Network learns exactly which optimizations work best for your specific hardware. Safety is built in. Every registry change requires Human-in-the-Loop approval — the AI explains its reasoning via a local Qwen 2.5 language model, and the user must click Approve or Reject. A one-click Emergency Rollback restores all settings to factory defaults. The dashboard includes real-time GPU frametime charts, an Active Target Lock that detects your running game and boosts it instantly, a Live Synapse Terminal showing AI decisions, Neural Net Q-Value visualization, manual overrides, and demo tools with simulate/stop stutter controls. Runs 100% locally, zero cloud dependencies, auto-admin elevation. Just double-click and let the AI learn your PC.
10 May 2026
