ReSim, the test infrastructure platform for physical AI, is now SignalFlag. Why the name changed, and what we are building next.
When we first launched ReSim in 2022, we were focused on building orchestration tools for robotics simulation, so our original brand name fit quite well.
Since then, teams have moved from pilots to fleets, and from behavior they wrote to behavior they train. Testing moved with them: onto hardware benches, into field experiment runs, and increasingly into production data.
A new name
No matter the modality, what teams need from us is the same every time:
Signal on how their robots are actually performing. Problems flagged early enough to act on.
Unless you’re a nautical expert, and none of us were before this exercise, signal flags are how ships communicate the state they are in, at a distance and at a glance. That is our job, so that is the name. ReSim is now SignalFlag.
What else changed this year?
Something else has changed, and it is the reason the new name comes with a shift in what we are building.
Generating code is getting cheap. Verifying it is not.
As the cost of generation falls, verification becomes the expensive half of the development loop. In robotics that half is slower and more expensive than anywhere else because evaluation is tied to physical hardware and physical time. So improvements to your verification loop are worth progressively more, not less.
At the same time, coding agents have changed the economics of tooling development. Teams could always build internal tooling shaped exactly to their workflow, but it was rarely worth the engineering time. Now it frequently is, and increasingly teams are taking this path, but there is a gap between working and production-quality, and closing it still takes iterations and breakages.
Some things your agent cannot generate for you, however many tokens you spend: accumulated test history, reliable operation, orchestration across thousands of jobs. Those are properties of a system that has been running for a long time, not properties of code.
So the interesting question is no longer build or buy. It is what you build on.
SignalFlag is a test infrastructure platform for physical AI:
- An MCP, which we now treat as the primary interface. Point your coding agent at it and build your test infrastructure from components already proven in production.
- A web app your agent can configure, so results persist and they can be reviewed, compared, and shared over time.
- A CLI and API for managing tests programmatically.
- Integrations with GitHub Actions, GitLab, VS Code, Slack, Foxglove MCAP viewing, S3 and GCS.
- Forward deployed engineers, when you have the foundation, but you want the additional confidence and speed that comes with a robotics infrastructure expert.
On the horizon
But we’re not just shipping a new brand name…
Self-serve access goes live today. Create an account, connect your agent to the MCP, and go. More MCP coverage and session evaluation are close behind, and both land in the coming weeks.
The fastest way to evaluate this is to hand it to your agent and see what it builds. Start free today.