Tomko & Company
Software is being reinvented.
AI is changing what software can do, how it’s built, and how people interact with it. We think that creates an opportunity to design new products, not just make existing ones faster or cheaper.
Tomko & Company builds those products. Our work spans personal computing, finance and engineering, and we’re developing the engineering methods needed to build them well.
Work
Civil
Civil takes engineering drawings and turns them into information an engineer can work with. We’re starting with utility engineering and building its capabilities one at a time.
We want to see whether we can build a junior civil engineer through software. That means teaching Civil individual competencies, testing them against real work, and making sure an engineer can understand and check what it produces.
Penny
Penny brings our household’s financial records together in one place. It starts with investment statements: each month we add the new ones, and Penny updates our household ledger, keeps the source of each printed value and exchange rate it uses, and marks a month incomplete when anything is missing.
We’re building toward a system that understands a household’s finances over time. We want people to be able to ask what changed, understand why, and trust that every answer can be traced back to their own records.
Haven
Haven is our local AI system for the home. Haven 1 combined a physical device, an iOS application, family photograph synchronization, memory retrieval, voice interaction and a television experience.
We want to rethink personal computing around the home: software that understands the information and devices in a household, works naturally across them, and keeps people in control of their own data.
Mila
We’re developing Mila to organize our family’s photographs and help us find memories through the people, places and events they capture. We’re starting with a locally stored archive and chronology.
We take thousands of photographs, but finding the one we remember is often harder than taking it. We want families to be able to find and revisit their memories without having to remember when a photograph was taken or which album contains it.
Alfred
Alfred keeps track of the devices its network observations reach and reports what changed between observations.
We’re building toward a home assistant that understands its environment and can eventually take action on your behalf, with your permission. We want software that helps run the home without requiring you to constantly open an app or give it instructions.
Roger
Roger is the internal engineering system we use to build our products. In a June 2026 benchmark, it took Civil 0.1.3 stories through task drafting and AI review. That benchmark did not preserve fidelity.
We’re rethinking how software is built. As frontier models improve, we want to take advantage of what they can do without losing control of the engineering process. We want Roger to keep the work organized, tested, and available to the next agent or model that picks it up.
HavenOS
HavenOS is the local software platform we’re building beneath our products. We’re developing a common runtime for agents, services and the devices they work with.
We want software to carry its history forward: what happened, what was decided and what remains to be done. That continuity should survive across sessions, devices, agents and models, so people don’t have to keep starting over.
Haven Engineering
Haven Engineering is the engineering standard we use to build software at Tomko & Company. It defines how we specify work, implement it with AI agents, validate the results, and decide what is ready to release.
We want to build serious software with AI without losing control of the engineering process. That means knowing what was requested, what was built, what was tested, and what evidence supports releasing it—even when different agents and models do the work.
Selected releases
Penny
Penny learned to keep a household investment ledger current: a new statement enters when it is added, and a month counts as complete only when every account and exchange rate is there.
Civil
Civil learned to treat notes and standards citations as records of their own, and to show what it could not resolve.
Alfred
Alfred learned to build an inventory of the devices its network observations reach, and to report what appeared or changed between them.
Penny
Penny learned to reconcile investment records across institutions and currencies into one household view, while preserving the source of every converted value.
Haven
Haven 1 brought a local device, iOS application, family photograph synchronization, AI-assisted memory retrieval, voice interaction and television photo display together into a working system.
How we build
Competence
We build with people who know the work. We teach the system one part at a time, test it against the real job, and improve it.
Continuity
Software should remember what happened, where information came from, and what changed. That is what HavenOS is for.
Sovereignty
Good software should work for the people using it. We prefer private and local systems when they make sense, and we want people to stay in control of their data and tools.