Zen Mascot

I build simple solutions
for complex problems.

Hi, I'm Ghazi. I build lightweight tools for developers, focused on performance, clarity, and privacy-first systems.

See my projects
file_type_python file_type_mysql
Currently Building

docmd

Build production-ready documentation from Markdown in seconds. No React, no bloat, just content.

DocumentationMarkdownCLIBunNode.jsTypeScript
Check it out →
docmd

memoryblock

Private AI workspaces that run locally, using fewer tokens and less compute.

AILLMPrivateLocalWeb AppCLILibraryBunTypeScript

AIMonitor

Native macOS menu bar app for monitoring AI usage, quotas, and provider status in real time.

macOSSwiftAIMenu BarOpen Source

Latest Rollups

View all →
Sep 9, 2026

HydraFusion: How GitHub Copilot Now Routes Between Multiple AI Models

GitHub's new HydraFusion research preview intelligently orchestrates multiple AI models to balance cost, performance, and quality in coding tasks.

Sep 9, 2026

GPT-6 Astra on Amazon Bedrock: What This Means for Enterprise AI

OpenAI's GPT-6 Astra is now generally available on Amazon Bedrock. I break down what this launch means for developers building production AI applications.

Sep 8, 2026

NeoMME: The Efficient Multimodal Encoder That Skips the VLM Bloat

NeoMME ditches separate vision towers for unified multimodal encoding. I break down why this matters for document retrieval and what it means for building efficient AI systems.

Sep 8, 2026

NeoMME: Building Document Retrieval Without Pretrained VLM Overhead

NeoMME is a 260M/800M multimodal encoder that rethinks visual document retrieval by training a single bidirectional Transformer from scratch instead of stacking pretrained components.

Sep 8, 2026

HydraFusion Shows How Multi-Model Routing Changes Developer Economics

GitHub's new HydraFusion research preview uses runtime orchestration to route tasks across AI models, delivering frontier-level coding quality at 67% lower cost than single powerful models.

Sep 7, 2026

BZip3: Why Compression Algorithms Still Matter in 2024

BZip3 achieves better compression ratios than its predecessor through advanced entropy coding and Burrows-Wheeler transforms. What this means for your infrastructure.