DISCOVER
Research agents propose candidate control laws for uncertain, nonlinear dynamics and try to break them.
AI-DRIVEN AEROSPACE ENGINEERING
Aerospace engineering,
driven by AI.
We build engineering tools that AI assistants operate directly, so an engineer and an assistant can do real analysis and design together. Alongside them, two research programs ask what those assistants can discover: control laws that hold up under test, and processors designed to run them.
FLIGHT TOOLS AVAILABLE · RESEARCH IN FOUNDING PHASE
FIG. 00 / PROGRAM MAPHOW THE PROGRAMS CONNECT
AVAILABLE NOW
Flight tools · 70 MCP tools
IN RESEARCH
Control discovery · Silicon
METHOD
Hypothesis → experiment → evidence
THE PROGRAMSOne toolchain.
One toolchain.
Two research programs.
The flight tools are available now. The two research programs are in their founding phase, so their pages describe aims, not results.
01 / FLIGHT TOOLSAVAILABLE
Flight dynamics and control, driven by your AI
Trim, linearize, simulate and design control laws for an aircraft through the Model Context Protocol. Written in Rust, with no MATLAB or Simulink.
EXPLORE THE TOOLS →02 / CONTROL DISCOVERYFOUNDING PHASE
Autonomous control-law discovery
Can long-running AI research agents discover control laws that beat a strong, fairly tuned baseline on conditions they have never seen?
EXPLORE THE PROGRAM →03 / SILICONFOUNDING PHASE
Autonomous processor design
Can AI agents design, verify and improve specialized processors for deterministic, low-power control, from RTL toward real silicon?
EXPLORE THE PROGRAM →HOW THEY CONNECT
Discover the control law. Prove it flies. Design the silicon.
The long-term aim is one engineering loop: research agents propose control laws, the flight tools test them against the evidence an engineer would demand, and the silicon program designs hardware to run the ones that survive.
PROVE
Candidates fly in nonlinear simulation with turbulence, across Monte Carlo dispersions, and are rated against MIL-F-8785C flying-qualities criteria.
IMPLEMENT
The silicon program explores fixed-latency, low-power processors to execute the control laws that pass.
THE STANDARD
Real engineering. Measurable evidence.
The same rule applies across every program, whether the engineer is a person or an AI.
An engineering idea from an AI is a hypothesis until an independent test says otherwise.
We treat AI-generated engineering ideas as hypotheses, not accomplishments. The work demands explicit assumptions, independent tests, reproducible artifacts, and comparison with strong conventional methods. No performance claim is published without a traceable experiment.