What I ship now

I don't talk about AI. I run it in production.

Not slideware. Live systems with Claude in the decision loop, real users, real SMS, and compliance constraints — built and operated solo. More production infrastructure than I had at a funded dot-com in 2000.

01 Lead artifact

Proof of Human Work (PHW)

Prove your work. Own your proof. Trust no platform.

A public protocol for the deepfake era. The thesis: machine output is now the background; human work is the signal — and that signal needs cryptographic verification. Three guarantees — Privacy, Truth, Proof: raw data stays encrypted locally, assertions are signed to evidence you control, and any verifier confirms offline with no platform in the middle. An Identity Layer (the Vault) plus a Filter Layer (consent); the first credential is Proof of Acquisition. This is the public face of the deepfake-breaker thesis.

instockornot.club/the_vault ↗

The Castaways — a fleet, and the blog that is its spine

A fleet of Claude agents across several machines, coordinating through a shared internal blog that is the spine of the org. Internal-first, always. Interesting stories get promoted to the public log — and one Promote button is the only thing that talks to the outside world, behind a preflight sanitizer that redacts internal addresses and vault paths.

It's not ‘I use AI.’ It's a fleet running under a contract I wrote, in production, every day.

The Skipper — public log ↗

Drop Watcher

Production collector-alert platform, live and paying its own way. Claude in the decision loop, carrier-approved SMS, and a real test suite — the full stack, solo.

Stack: Python · Flask · SQLite (WAL)
Messaging: Twilio A2P 10DLC
Tests: 43-test suite
Status: Live · real users

github.com/simonhg321/drop-watcher ↗ instockornot.club ↗

The Billboard

The domain-agnostic successor to Drop Watcher. Same Claude-in-the-loop architecture, anti-sticky by design, scaled across many sources for the price of a coffee.

Sources: 94+ sites
Architecture: Claude-in-the-loop
Run cost: ~$0.14 / week

The substrate

Beneath PHW sits an encrypted Vault — a local store on a Mac Mini with a local model and Fernet authenticated encryption at rest. Plaintext metadata for search, ciphertext for every title and body. Nothing leaves the network. It's the identity substrate the public protocol is built on, not its headline.

Private builds I run

The visibility career, run at home.

Three internal tools on the home network — shown as sanitized screenshots, never linked. Captions describe capability, never internal addresses.

Internal command-center dashboard showing fleet metrics — The Interface The org's internal command center: live fleet status, metrics, and the internal blog that gets promoted to the public log.

Internal blog editor with Promote control and a promoted post — The Skipper — internal → public Internal-first writing. One Promote button is the only thing that talks outward, behind a preflight sanitizer.

Home-network visibility dashboard with VLAN distribution — The Home — network visibility L2–L8 visibility at household scale: device inventory across VLANs, WAN-block monitoring, hundreds of thousands of syslog lines.

All Knowing

The one that isn't pictured: AI log forensics. A SQLite index over ~1.7M log lines with sub-second queries, source/tag/date filters, and a live tail over SSE. A local-model layer turns ?-prefixed plain English into SQL (shown to you), and an Analyze button returns a security-analyst writeup citing timestamps and IPs. The visible heir to the Bloomberg/Gigamon network-visibility work — run at home, on a local model.