The Lab · v0.2

My projects.

The deep-dive companion to niclydon.com. Where the front door is a portrait, this is the schematic — eight projects, seven agents, and the home lab they all run on. Two AMD Strix Halo machines linked over Thunderbolt 5, an M4 Mac Mini for dev work, eight always-on inference models, one ~191-table Postgres brain, zero cloud dependency for anything I actually care about.

• 01
  ARIA Adaptive Responsive Intelligence Assistant
  Next.js · TypeScript · Swift 6
  Active
  →
• 02
  Desk Operator console for the Nexus agent platform
  Next.js 16 · React 19 · TypeScript
  Active
  →
• 03
  Nexus Unified agent platform
  TypeScript · Express · PostgreSQL
  Active
  →
• 04
  Forge Home lab LLM inference gateway
  FastAPI · Python · llama.cpp
  Active
  →
• 05
  Homelab Host-level infrastructure and model lifecycle contract
  Python · Bash · systemd
  Active
  →
• 06
  Tongs Unrestricted permissive chat with visible thinking
  Next.js 16 · React 19 · TypeScript
  Active
  →
• 07
  Whittled Whittle your photo library down to what matters
  Swift 6 · SwiftUI · PhotoKit
  Pre-launch
  →
• 08
  Voice Print Media processing & person fingerprinting hub
  Python · Next.js · Whisper
  Active
  →
• 09
  Cairn Daily check-in capture surface for journaling, biography, and drafts
  Swift 6 · SwiftUI · PostgreSQL
  Experimental
  →
• 10
  QLoRA Personal language model fine-tuning
  Python · Unsloth · QLoRA
  Experimental
  →
• 11
  Broadside Multi-brand editorial desk seeded from your own project docs
  Next.js 15 · React 19 · TypeScript
  Active
  →
• 12
  Jingle Family Multi-tenant Christmas elf universe
  Next.js · TypeScript · Tailwind
  Active
  →
• 13
  Smithy Image & video generation workbench
  FastAPI · React · TypeScript
  Experimental
  →
• 14
  Netware Blue Retro design system inspired by 1990s Novell NetWare
  TypeScript · Node.js · ESM
  Active
  →
• 15
  ARIA Origin How a chatbot became a personal operating system
  Next.js · PostgreSQL · Swift
  Archived
  →

The substrate that makes the projects above possible. Two AMD Strix Halo machines (Furnace + Crucible) linked over Thunderbolt 5, an M4 Mac Mini for daily dev (Anvil), and a remote KVM (Bellows). Combined: 192GB RAM, 160GB VRAM, eight always-on inference models, zero cloud dependency for personal workloads.