Exemplar is part of the NVIDIA Inception program.
NVIDIA InceptionExemplar is part of the Google for Startups program.
Google for StartupsConfigurable autonomy
A software factory is how the organization allocates unsupervised agent work: which jobs an agent may take, how far it may go before a person decides, and how the org proves what happened. Configurable autonomy for engineering and product — path by path — not a faster coding queue.
Not a bigger model. Not a faster IDE. Not a queue of generated pull requests. A production line has named stations, a refusal cheaper than scrap, and someone who still owns the part after it ships. The agent is labor on that floor. The factory is the floor.
Engineering and product already feed the same production. If only engineers have a dimmer, the rest of the company is still shipping agent output — specs, copy, tickets, launch checklists — without a named owner for refusal.
Illustrative — autonomy as a budget, path by path.
An agent may rewrite internal lint without asking. The same agent needs approval before it touches a billing contract. Configurable autonomy is that difference, encoded as policy — not as culture.
A PM uses an agent to draft a launch checklist that later becomes a production change. If the factory only covers the IDE, that work never hits allow / ask / deny. The factory has to cover both surfaces.
Security asks what shipped last week. A working factory can narrate it, interrupt a path that drifted, and name the owner. Volume of generated diffs is not the metric.
Applied installs a software factory harness when that is how you ship. The lifecycle underneath is ADLC. Longer argument: A software factory is configurable autonomy.
Schedule a briefing →