FAQ
A few direct answers before the Factory Sketch.
What does Singularity Labs actually build?
Custom agent-operated workflows for expensive knowledge work: intake, planning, role assignment, review, memory, logs, handoffs, recovery, and the human operating loop around it.
Is this an autonomous agent platform?
No. The point is not to pretend judgment disappeared. The point is to design the factory so the right work moves through agents and the right human sees the right risk at the right time.
What is a Factory Sketch?
A focused diagnostic engagement. We map one workflow, identify what can safely move to agents, define the foreman role, list the risks, and decide whether a build makes sense.
What is foreman training?
Training for the person who will operate the factory after handoff: writing usable specs, reading handoffs, spotting drift, approving risk, improving memory, and stopping the line when quality drops.
What happens when the agents break?
Assume they will. The factory needs checks, logs, recovery paths, and escalation rules so failures become visible and useful instead of mysterious.
Can Singularity Labs host and operate the factory?
Yes. For clients who want that, we can run the factory on managed infrastructure with a private dashboard, controlled agent access, usage logs, backups, updates, and recurring billing for hosting and upkeep.
Is Factory Control the same thing as a managed factory dashboard?
No. Factory Control is the scoped workbench for managed site and agent work. A managed factory dashboard is a deeper client portal surface for a dedicated factory: host status, upkeep, feedback events, blocked decisions, and foreman handoffs.
Do clients get AgentOS Factory?
AgentOS Factory is the demo path and internal proof target. A client gets a factory designed around their own workflow, hosted and exposed only as much as the engagement requires.
What does the AI usage cost?
There are two different costs. Build-time experimentation is part of research, testing, and tuning the factory. Production usage is measured after the decision loop is stable enough to benchmark, then quoted as pass-through usage, a monthly allowance, or a capped operating budget.
Do you publish the thinking behind this?
Yes. The field notes stay canonical on davidmieloch.com, and Singularity Labs links to the relevant pieces instead of maintaining duplicate article copies.