News
Building AeroGenie.
A running log of how we are building AeroGenie's agentic AI decision engine — the runtime, the AI simulations, the multi-agent AI, and the governance underneath.
June 2026
Early access opens.
We opened the AeroGenie waitlist and began onboarding a first group of design partners across aviation, finance, supply chain, and security. Early access focuses on real, high-stakes decisions — pricing moves, disruption response, cash forecasting, and incident triage — run end to end inside one governed workspace.
May 2026
Mojo-accelerated runtime now sustains thousands of what-if simulations per second on a single node.
Decisions as probability, not point predictions.
The simulation engine now returns full outcome distributions for a decision — lower bound, upper bound, the most likely result, and the downside risk — instead of a single number. Teams can stress-test assumptions, compare paths, and see which variables actually move the outcome before they commit.
MCP connectivity: attach databases, files, APIs, and tools to a mission with no custom glue code.
April 2026
Capability forge: agents assembled for the mission.
Rather than relying on a fixed library of prebuilt agents, AeroGenie now composes the agents a decision needs on the fly — one to read the data, one to run the simulations, one to weigh tradeoffs, one to prepare the decision pack. The mission runtime coordinates them into a single, reviewable plan.
Replayable audit trail: every mission can be replayed step by step — data, assumptions, approvals, and actions.
Bring your own LLM — the provider router now supports Anthropic, OpenAI, Google, and local models.
March 2026
Triggered decisions.
Missions no longer have to start with a prompt. AeroGenie can now ingest an event, alert, metric threshold, or system signal and start the workflow automatically — moving from signal, to analysis, to decision, to action, with a human in the loop wherever policy requires it.
Human-in-the-loop approvals with policy-based autonomy — autonomous where it is safe, controlled where it matters.
February 2026
The mission workspace.
A first end-to-end look at the workspace — a Cursor- or Codex-style environment, but for business operations. A mission starts from an objective, finds the right context and data, builds a path from question to decision, and keeps every step visible and traceable back to evidence.
Full-context analysis: read entire documents and datasets instead of sampling rows or skimming pages.
January 2026
Simulation + ML: simulations now stress-test and enrich machine-learning outputs with probability ranges.
First Mojo-accelerated function library: optimization, forecasting, numerical analysis, and uncertainty modeling.
December 2025
A scientific-grade decision runtime.
We brought up the engine beneath the chat layer: a high-performance mathematical runtime, accelerated with Mojo, built for simulation, optimization, probability, and forecasting. LLMs handle intent and interaction; the heavy reasoning runs on optimized math, grounded in data and audit-ready.
Published our explainer film on turning enterprise data into clear, governed decisions.
November 2025
Internal note: why simulation scales across many questions where a single trained model narrows.
October 2025
It began in aerospace.
IAG brought us a decision problem from one of the most demanding operating environments in business: aviation, where aircraft, routes, weather, crew, fuel, and safety all interact. It needed more than a chatbot, a dashboard, or a spreadsheet. We built AeroGenie for that world — then realized the same pattern repeats wherever high-stakes decisions are made.
AeroGenie founded — to build an agentic decision engine for complex, high-stakes work.