We're an AI-first studio without a wall of client logos, and we won't invent one. What we have instead: our own production systems — built exactly the way we build for clients, running right now, with the code public wherever it can be. Every claim on this page is inspectable.
The problem. Our old site was a 2019-era jQuery/Bootstrap template — heavy, generic, and nothing like the engineering we sell. A studio that pitches lean, modern software can't hide behind a template.
A complete rebuild in a single day (31 August 2026): hand-written HTML, CSS and vanilla JavaScript. Zero frameworks, zero build step, no trackers, no cookies, no analytics scripts. AI-assisted throughout — every decision made by a human.
The entire stylesheet is about 13 KB. Pages ship exactly as written — no bundler, no transpiler, nothing between the repo and your browser. Every push to the public GitHub repo deploys automatically via CI to Firebase Hosting.
Lean engineering is a choice, not a limitation — and AI-assisted development is fast when humans stay in charge of the decisions. The site you're reading is the evidence, and its source is one click away.
Don't take our word for it — read the site's source on GitHub →
The problem. A contact form that dumps messages into an inbox — or worse, a third-party form SaaS — means manual copy-paste, slow replies, and someone else holding your leads. Exactly the kind of small, dull friction we tell clients to automate away.
An end-to-end pipeline we own outright: the contact form writes each enquiry to a
Firestore leads collection, the visitor instantly gets a reference number
on screen, and an automatic confirmation lands in their inbox. No form SaaS anywhere
in the chain.
Firestore security rules are create-only — the public can add a lead, never read or alter one. Each lead gets a 10-character reference number shown to the visitor. A Cloud Function fires on the Firestore write and sends the auto-reply over SMTP with nodemailer, quoting that reference.
Small, reliable automation that removes manual work — designed with least-access security from the first line. This is the shape of most first projects we build for clients: unglamorous, wired into what already exists, and it just runs.
Want to see it in action? Send us a message via the contact form and watch your reference number arrive.
The problem. Businesses run on Java Swing desktop applications that most agencies write off as untouchable — no API, no web interface, no way for an AI assistant to help. That software isn't going anywhere; the tooling had to come to it.
Swing MCP: an open-source MCP server that lets AI assistants operate Java Swing apps directly — take UI snapshots, click, type, fill forms, read tables and trees, drive menus and dialogs, capture screenshots. Now at v1.1.0 and listed in the MCP registry.
A multi-module Maven project on Java 21 and Spring Boot. A Java agent attaches to the target application and exposes its live UI over MCP — the open standard, originated at Anthropic in 2024 and since adopted across the AI industry. Installable as an MCPB bundle.
We build AI connectors for the software everyone else calls impossible — and we do it in the open, where anyone can audit the code, file an issue, or run it against their own systems before ever talking to us.
It's all public — read the Swing MCP overview or explore the code on GitHub →
Start with a free AI Opportunity Audit: a 30-minute call, then a written automation map for your business within a week. If AI isn't worth it for you, we'll say so — honesty is the whole point of this page.
Book your free audit