Our work
What we've built. What we're exploring.
Client work
Land finance: built around the finance team
In production · AI-assisted development
A homebuilder needed a custom application to manage lot purchases. We designed and built a custom application, with AI assisting throughout development. It went from concept to production in 3 months and is now part of the finance team's day-to-day operations.
Invoice validation: find savings in the details
Validation
Helping business teams find avoidable costs in the invoices they handle. We developed an automated invoice-validation process and advised on which AI projects to pursue, how to build them, and how to oversee their use. Invoice validation is designed to reduce avoidable spending. The wider engagement also strengthened the business's practices for developing and overseeing AI.
Subcontractor data: a better home for business data
A design-build contractor relied on spreadsheet databases to manage subcontractor information and support bid preparation. We built a shared platform for subcontractor records, with administration tools, controlled access, and bid sheets populated directly from those records. The team can keep subcontractor details up to date and prepare bids from the same records.
Experiments
Working projects that help us test what AI can do. These are experiments and prototypes, not client deliveries.
- LightGarden (working experiment): Lumen, the AI artist behind LightGarden, creates art for a 16 × 16 LED canvas (256 pixels), reviews its drafts, and records what to try differently next time. Each finished piece comes with its drafts, feedback, and a journal entry.
- Dipu (prototype): a meeting agent that takes action on requests during live meetings, working alongside the people in the room.
- Salon operations (internal experiment): we're testing agents in our own salons on the work that usually happens between appointments, across marketing, HR, and finance.
- CHAI (working prototype): a watch shortcut that captures spoken ideas in Reminders and syncs them to the Mac.
Our next chapter
We're building CodeTicks around people and agents, with clear responsibilities for both. People set direction, make decisions, and own the outcome. AI agents plan the work, use tools, and deliver results.