Director of IT · Atlanta, GA · lasswell.me
I lead IT infrastructure and service delivery — data center, cloud, security, and the operating model that holds them together. Twenty years in, most of the work is organizational: deciding what to standardize, what to retire, and who owns the thing at 3 a.m.
I also never stopped writing code. The Home Assistant integration further down this page runs in roughly 5,400 homes. Keeping something in production that strangers depend on is the cheapest way I know to stay honest about what I ask engineers to do.
Infrastructure is a product. Core systems earn the same rigor as software — roadmaps, service levels, versioning, a named owner. The moment the network has a backlog instead of a ticket queue, the business starts talking about it differently.
Governance should accelerate decisions, not queue them. Most change advisory boards are a tax on velocity that buys no safety. Guiding principles used as decision filters push judgment down to the people holding the context.
The service desk is the product surface. It is where the entire IT organization is actually evaluated. Staffing it like a cost center and then wondering why IT has no credibility is a self-inflicted wound.
Ship it. Perfect is the enemy of shipped, and a platform nobody uses is a rounding error no matter how elegant the architecture.
A ten-part series on modernizing ITSM for cloud-native, multi-framework organizations:
- Why ITSM Still Matters in a Cloud-Native, Agile, Multi-Framework World
- The Service Value System: Connecting Strategy to Execution
- Governance Without Gridlock: Agility and Accountability in Service Delivery
- Infrastructure as a Product: Treating Core Systems With the Rigor of Software
- Change Enablement in a DevOps World: Replacing Fear With Velocity and Trust
- Service Portfolios & Productization: Defining and Managing IT Services
Earlier series cover technology leadership, tech roadmaps, and the outsourcing versus in-housing decision. Full archive at lasswell.me.
Leadership is the day job. This is the part that keeps me fluent.
Two HACS integrations, both written against cloud APIs that were never documented for this purpose. Between them they cover most of what is plugged in at my house.
A HACS custom component covering Govee lights, plugs, fans, humidifiers, heaters, sensors, and leak hubs. Capability-based rather than SKU-based — entities are built from what each device advertises, so new models in a known class generally work without a release. Real-time push over Govee's AWS IoT MQTT, with automatic local LAN control and cloud fallback. Python, MIT, released most weeks, and by a wide margin the most-used thing I have shipped.
Accepted into the HACS default registry in January 2026, so it installs from inside HACS without adding a custom repository.
→ lasswellt/govee-homeassistant
Tankless water heaters and combi-boilers over the NaviLink cloud. The interesting
part is the client: REST authentication, AWS SigV4 WebSocket signing and an MQTT
transport, written from scratch and asyncio-native, with no boto3 and no
AWSIoTPythonSDK — paho-mqtt driven on the event loop with no background network
thread. Telemetry is capability-gated, so a unit only grows the entities it actually
has.
→ lasswellt/navien-homeassistant
I run Omarchy as my daily driver and write bar plugins for it. Both of these answer the same question — how much of my AI quota is left — for tools that do not otherwise say.
Each is a service plus a bar-widget: a collector emits one JSON usage record in
the same contract Omarchy's first-party collectors use, a QML service publishes it on
a timer, and both the built-in Agents panel and the plugin's own panel read that one
file. One data path, two views, so they cannot disagree.
| Plugin | What it surfaces |
|---|---|
| omarchy-copilot | GitHub Copilot premium-request quota with a reset countdown, AI credits, tokens by day and model, and which repositories the work happened in. Quota is read live from the CLI's JSON-RPC runtime rather than its cache, and costs no premium requests to collect. |
| omarchy-antigravity | Google Antigravity limits across all four quota buckets, merged across the agy CLI and the IDE because the quota is account-wide. Ships a smoke suite that fails loudly if Antigravity's JSON envelope changes shape. |
Both MIT, both QML with shell and Python collectors.
| Project | What it is | Built with |
|---|---|---|
| blitz-cc | A language-agnostic agentic development loop for Claude Code — sprint workflow, agents, quality gates | Shell, Markdown |
| cc-metrics | Self-hosted OpenTelemetry dashboard for Claude Code token use and cost | JavaScript, OTel |
| claudeHQ | Workforce management for Claude Code: monitor and control sessions across machines from one dashboard | Nuxt 3, Vuetify, TypeScript |
| signalslate | One morning page — mail, calendar, tasks and chat across every account, rendered to PDF and pushed to a reMarkable | Python, Docker |
| flight-search | Flight search on the Amadeus API — Vue frontend, Express backend, and an MCP server so assistants can query it directly | Vue 3, Express, MCP |
| remodel-planner | Room-by-room remodel planner: SVG floorplan with snapping, phase-gated tasks, budgets, permits | TypeScript, Firebase |
Two platforms are in private development: CubeSP, a multi-portal ITSM/PSA platform for MSPs, and MEMBRIX, a membership and event operating system. Both are TypeScript monorepos on Vue 3 and Firebase.
Architecture notes live in playbook-library, a wiki-only repository. Its centerpiece is a multi-project Quasar and Firebase playbook — several independently deployed apps sharing one Firebase backend, composed through module federation — with three runnable companion repos as its worked example: host, knowledge and administration.
Azure and AWS, Kubernetes and Docker, Firebase and Firestore. Python for integrations and automation, TypeScript and Vue 3 for products, PowerShell where the fleet lives. Datadog and Microsoft Sentinel for observability, Azure DevOps for pipelines, SQL throughout — plus a working knowledge of every ITIL practice I have opinions about.
LinkedIn · lasswell.me · Medium · X