All Series Archives

Building Maabarium

Part 1 of the Building Maabarium series: the origin, architecture, and motivation behind an autonomous AI research laboratory built for continuous, visible experimentation.

Part 1: Maabarium: Building an Autonomous AI Research Laboratory That Never Sleeps Part 2: Maabarium v0.4.1: Refining the Engine of Discovery Part 3: Maabarium: Blueprints, Evaluators, and Experimental Design Part 4: Maabarium: Councils, Git Worktrees, and Safe Execution Part 5: Maabarium: Running a Local-First AI Lab in Practice

Start Reading →

Building PortForge

Part 1 of the Building PortForge series: why PortForge exists and how it brings framework, Git, container, and health context to local port inspection.

Part 1: PortForge Part 1: Engineering Context for Local Port Inspection Part 2: PortForge Part 2: Proactive Discovery, Deeper Insight, and a Terminal Interface That Feels Alive

Start Reading →

Building Mnemara

Part 1 of the Building Mnemara series: why Mnemara exists, how its explainable episodic memory works, and why durable memory is essential for trustworthy AI agents.

Part 1: Mnemara: Giving AI Agents a Memory They Can Truly Trust Part 2: Mnemara v0.5.0: Time-Travel, Judged Evaluation, and Changefeeds

Start Reading →

Building Vona

Introducing Vona, an open-source Rust runtime for production voice AI, with provider-neutral speech backends, deterministic tests, and interruption-aware sessions.

Part 1: Vona: The Open-Source Rust Runtime That Powers Production Voice-Native AI Part 2: Vona v0.2.0: Local Voice on Apple Silicon & Ollama Streaming

Start Reading →

Building Nexara

Introducing Nexara, the capability layer for AI systems that need signed skill discovery, policy-bound tool calls, remote execution, redacted secrets, audit trails, and optional learned ranking without hard-coding every integration.

Part 1: Nexara: Safe Tools for Agentic Products

Start Reading →

Building PermeantOS

Introducing PermeantOS and USXF, an open-source research preview for migrating live AI runtime state across heterogeneous inference engines.

Part 1: PermeantOS: Portable State for Migratory AI Agents Part 2: PermeantOS: Agents Can Move

Start Reading →

Building QATQ

The original QATQ research spark: combining quaternion structure with TurboQuant-style KV cache compression to ask whether local long-context inference could become dramatically more practical.

Part 1: Quaternion-Augmented TurboQuant: A Speculative Path to Extreme KV Cache Compression Part 2: QATQ: From Speculative Paper to Production Release

Start Reading →