Ian Provencher
Listen to the podcast

Architectures · 10

The systems I run

Interactive diagrams — click any node to inspect it, or step through the walkthrough. Rendered from public-safe records exported out of my Command Center.

10
systems
80
nodes
99
connections

How a private system reaches this page, public-safe

Command Center OS

Local-first operations console where the UI and the AI work the same files — facts derive from live sources, and the console alone moves work to done.

10 nodes · 6 steps Explore →

AppliedIQ Solutions — the public face

A 17-page consulting site with live supply-chain demos and a streaming AI assistant, all on Cloudflare's edge.

6 nodes · 5 steps Explore →

Nightly ops — plain code, honest exit

A 06:30 scheduled task that checks the build, syncs remote token numbers, commits the data, and writes a health record LAST — with an exit code that finally measures something.

7 nodes · 5 steps Explore →

The brain snapshot pipeline

How 803 private vault notes become a public-safe living graph — sanitized at the source, simulated in the browser.

6 nodes · 5 steps Explore →

'Ask the tooling' — the site's live assistant

A public AI assistant in Ian's locked voice — streaming, spend-capped, and adversarially hardened.

5 nodes · 5 steps Explore →

The mirror/bridge deploy pipeline

Claude artifacts become deployed full-stack apps through two composable skill prompts — then move to durable infrastructure.

6 nodes · 5 steps Explore →

Hermes — the shared brain

One knowledge vault, served to every AI surface over MCP, curated nightly by an autonomous librarian.

6 nodes · 5 steps Explore →

The demo box

One small server, many live demos — a mirror-first deploy contract with zero inbound ports.

6 nodes · 4 steps Explore →

Headless agents — the overnight fleet

Two autonomous Claude agents on a sandboxed box — sealed in containers, walled behind an allowlisted proxy, growing their own capabilities, and one of them holding a real wallet.

15 nodes · 9 steps Explore →

Model routing — the control plane

Every subagent used to inherit the biggest model at the highest reasoning effort, because nothing said otherwise. Three read-only file-mapping agents burned 9.88M cached tokens proving it. What the measurements found next was the opposite of the obvious answer — and the same instrument that justified the project is the one that ended it.

13 nodes · 7 steps Explore →

Data reviewed · current · refreshed from the private estate