Introducing Sim-1
Sim-1 is the first in a family of models trained to simulatehow code runs — not just predict the next token, but reason about program state, side effects, and failure modes the way an engineer does.
Why simulate?
Most coding models learn from text. Sim-1 learns from execution — tracing real programs across millions of runs, in real environments, and learning to predict outcomes before any code is shipped. The result is an agent that can verify its own fixes before a human ever sees them.
What changes for support engineering
With Sim-1, Supporty agents close the loop end-to-end: triage, reproduce in a private sandbox, generate a fix, simulate the patched system, and only then open a PR. The same memory that catches one regression catches the next.
Availability
Sim-1 is rolling out to Supporty enterprise tenants this quarter. Talk to us to get early access.