ALEXEI BROWN

OpenCode fork with working Anthropic OAuth and non-AVX CPU support

Maintained fork of OpenCode (Claude Code's TUI) that runs on non-AVX CPUs — Celerons and older accessibility hardware. Fixes Anthropic OAuth, improves the terminal UI and adds mobile optimisations.

Stack: TypeScript · Bun · OAuth 2.0 · CLI development

Gesture and voice layer on top of tmux for running multiple AI agents

Reimagining terminal multiplexing for AI-agent workflows: gesture-controlled agent dashboard, voice commands, tab cycling between tasks/chat/hardware, pinch-to-click overlay, 10 themes, live mock agent state.

Stack: TypeScript · MediaPipe · Terminal UI · WebGL

Single Signal + WhatsApp session exposed as a localhost HTTP API for AI agents

Tauri desktop app that consolidates signal-cli (Java) and whatsmeow (Go) into one local REST API on port 8769, so AI agents route all messaging through a single owned session instead of fighting over companion states. Health checks, send/read endpoints, append-only JSONL inbound log, sandboxed public-user vs authorised-user routing.

Stack: Tauri (Rust + TypeScript) · signal-cli · whatsmeow (Go) · REST API

Voice-controlled terminal interface for AI coding workflows

Speech-to-action layer for terminal-driven AI coding: spoken commands routed through Whisper into shell + agent invocations, including speaker-recognition for multi-user environments.

Stack: Python · Whisper · Speaker recognition · Linux

30B-param MoE LLM running 100% on an AMD iGPU — 38 tok/s, zero cloud

Deployed Z.ai's GLM-4.7-Flash (30B-A3B Mixture-of-Experts, Q4_K_M, ~19 GB) entirely on an AMD Radeon 8060S iGPU via ROCm 7.2. The default 198K-context model OOM'd (KV cache alone needed ~53 GiB vs 32 GiB VRAM); solved by engineering a custom 32K-context variant that brings total memory to ~28 GiB for 100% GPU execution with zero CPU offload. Integrated as the offline fallback for local coding agents.

Stack: Ollama · ROCm 7.2 · PyTorch · GLM-4.7-Flash (MoE) · Model quantisation

SDR (HackRF / Airspy / RTL-SDR) → real-time demod → GPU Whisper STT → searchable archive — fully local

Multi-layer software-defined-radio intelligence stack: captures RF across UHF/VHF/HF/marine/aviation bands, demodulates in real time (WFM/NFM/AM/SSB), transcribes voice traffic with faster-whisper on AMD ROCm, extracts entities (callsigns, locations) and archives everything into a searchable SQLite/FTS5 database. The DSP core was extracted into a standalone library (rfai-dsp) with anti-alias decimation, phase-continuous VFO mixing and an abstract SDR backend.

Stack: Python · PyQt5 · HackRF / Airspy / RTL-SDR (SoapySDR) · DSP (rfai-dsp) · faster-whisper + ROCm · SQLite/FTS5

Wearable ISR: glasses → home computer (face recognition) → phone — custom ONNX demographic models

Real-time computer-vision pipeline for smart-glasses use cases: MediaPipe Face Mesh (468 landmarks), ArcFace recognition, multi-face tracking at 30fps with a persistent identity library — all in-browser. Ships custom-trained 'Folkus' demographic models (2.3 MB INT8 ONNX, running via ONNX Runtime Web) that beat commercial APIs: gender 92.1%, age 5.37y MAE, ethnicity 72.2% across 7 groups. Camera glasses stream to a host computer that recognises faces and feeds AI context back via phone.

Stack: JavaScript · MediaPipe · ArcFace · ONNX Runtime Web · Custom ONNX models · WebGL