Building AI that can hold a continuing job
FriendMachine develops persistent AI operators: software agents designed to take responsibility for ongoing work while remaining subject to explicit limits, verifiable outcomes, and human direction.
The company develops these operators inside software products it owns. That creates a demanding proving ground where their work can be measured against real customers, operating systems, costs, failures, and commercial outcomes rather than demonstrations alone.
Why FriendMachine exists
AI models can already perform increasingly capable pieces of work. The harder problem is sustained delegation.
A company needs more than an agent that can complete a task. It needs to know what the agent is responsible for, what remains unfinished, what it is allowed to do, what actually happened, and when a person needs to intervene.
FriendMachine is building the organizational machinery around that problem.
A FriendMachine operator holds an ongoing role. Its assignment, approved authority, unresolved obligations, work history, and outcome record persist outside any individual model session. Reported results can be checked against the systems where the work occurred, and work that exceeds the operator’s authority returns to a person as a decision.
Building against real businesses
FriendMachine starts with its own products so the operator system has real consequences from the beginning.
Articul is the first product-operations proving ground, with ongoing work in distribution, customer support, measurement, and product improvement.
Geod provides a different operating environment centered on geospatial analysis and multi-location business decisions.
Other software, including OnlyBots, contributes infrastructure for persistent machine identity and organizational participation.
FriendMachine also maintains research programs in programming languages, agent evaluation, geospatial systems, and software for physical operations. Research is used both to improve the operator system and to explore technical questions that may matter beyond immediate product development.
Company and leadership
FriendMachine was founded by Joshua Winters, a software engineer and engineering leader whose work spans software products, data and geospatial systems, programming languages, and software for physical operations.
Before FriendMachine, he served as CTO of IdealSpot. Since 2020, he has led software engineering at Sensei Ag, working on systems that coordinate people, software, data, and automated physical operations.
What FriendMachine is trying to prove
The immediate objective is to demonstrate that persistent operators can assume meaningful continuing responsibility while reducing the amount of human supervision required, without obscuring uncertainty or lowering the quality of the work.
A fully autonomous company is a longer-term possibility, not the current claim.
Progress is published in the Build log, including the distinction between work that has merely been reported, independently verified, and observed to produce real downstream results.
Contact FriendMachine
FriendMachine works with software companies interested in developing a persistent operator for a specific product or operating role.
FriendMachine is interested in conversations with people working on agent infrastructure, governed AI systems, software operations, and potential environments for future persistent-operator deployments.
For questions about Articul, Geod, research, technical collaboration, or FriendMachine itself, use the contact form.