Wharf Machine is an independent software company and workshop. We make our own products: Mac-native, local-first, designed around AI from the beginning. Creative direction, interaction design and engineering are one job here, done by the same people. Alongside our products, we lead a small number of engagements for other organisations.
Currently taking a small number of engagements
01
We make tools we want to exist: native on the Mac, fast, and designed around AI from the beginning. You own what you buy. Intelligence runs locally where the task allows it, so capability does not disappear when a provider changes their terms. Recurring fees, where they exist, pay for recurring value.
02
Where a product's character gets decided: form, motion, tone of voice and the visual language around it. We work with generative image and video tools alongside conventional design tools, so direction shows up as rendered work early rather than as a deck. Brand and identity when the product needs one.
03
Native Mac software, web applications where the web is the right tool, and agentic systems, for organisations that want the same standard applied to their product. We lead the work from product direction and interaction design through prototypes, technical decisions and production software.
04
Helping organisations become AI-native across product, design, development and daily operations. We find where AI raises the pace and ambition of the work, then shape usable workflows and prototypes that teams adopt rather than admire, with evaluation loops that show whether a non-deterministic system is actually good enough to ship.
The part that usually decides whether the work is any good.
Direction, design and engineering are the same people. Nothing is translated at a handover, because there is no handover.
You see rendered direction and running software in the first weeks. Documents follow decisions rather than standing in for them.
We run few engagements at once, so the people who take the work do the work.
Code, weights, prompts and pipelines stay with you. No component of the system is ours to withhold.
It means keeping capability, data and economics close to the product.
Local AI / 03
Run intelligence on the user’s Mac, a nearby Linux server or small edge hardware whenever the task and hardware allow it.
Build on capable models that can be run, evaluated and replaced without handing the product to one provider.
Use hosted models only when their additional capability earns the additional operating cost.
Put AI inside clear, well-designed workflows, not beside the product as another generic chat box.
The values behind our own products, commissioned work and advisory.
Creative direction and interaction design decide what should be built and how it should feel. They are part of product direction, not a finish applied at the end.
We use generative models throughout the making, for imagery, for code, for evaluation. The work is judged on what it is, not on how it was produced.
Non-deterministic systems need evidence. If we cannot measure whether it is good, we do not claim that it is.
Native Mac where interaction quality, speed and platform fit decide the product; that is our default and our home. The web where reach and control matter more: a surface you open from anywhere, on anything, beats a better-feeling one you cannot get to. Chosen per job, never by ideology.
Prefer systems people can run, understand and keep. Recurring fees should pay for recurring value.