Over the last months, GoodAI has been running an internal project called Prometheus.
The idea was straightforward but ambitious:
build autonomous, intelligent drone swarms that can operate indoors, in GPS-denied and signal-denied environments, with no pilot on the sticks. You give them high-level goals, and they figure out how to execute.
We’ve now reached the point where this is no longer just a research experiment. We’re spinning Prometheus out into a dedicated effort called GoodAI Swarm Robotics, focused on turning this technology into a real product and company.
Before I go into more detail, here is a short demo of what we have so far:
Prometheus: What We Built in 4 Months
Prometheus started as a focused internal sprint at GoodAI. We pulled together a small team of four engineers, set a clear technical goal, defined our constraints, and gave ourselves only four months to see how far we could push the idea.
I’m genuinely proud of what the team achieved in such a short time. In those four months, they built two working Prometheus prototype drones and a separate prototype of a self-charging dock.
On the technical side, the current prototypes can already do quite a lot. They can fly indoors and close to obstacles using SLAM to map the environment and avoid collisions. Takeoff and landing are fully autonomous, without any joystick piloting. The drones are controlled by an agentic LLM running in the cloud today, connected over a low-latency link, which interprets high level instructions and reasons about what to do next. They can also find the charging dock, land, and recharge automatically.
There are many limitations left to solve. Flight time is short, and there are still tricky edge cases such as glass or very narrow corridors. We also need a lot more flight hours and stress testing. But for a small team and a four month sprint, this is a strong foundation.
What Prometheus Is Aiming For
The long-term direction for Prometheus is:
An autonomous swarm of drones, with full intelligence on board – like having a human pilot inside each drone.
Today, the LLM that controls the drone runs in the cloud. The architecture is already built around high-level goals, spatial reasoning, and agentic behavior.
The end goal is to move this intelligence fully onboard, so each drone can operate without any reliance on external compute or constant connectivity.
No FPV skills. No manual piloting. No dependency on a cloud backend that cuts out when the signal drops.
Instead, you give high level goals, such as:
-
Search the second floor and find any people
-
Map this building and highlight blocked exits
-
Inspect this structure and show anything that looks damaged or risky
The swarm then handles the rest:
-
Assigning drones to tasks
-
Coordinating and sharing maps
-
Returning to recharge
-
Merging the data into a navigable 3D map
-
Explaining the findings in natural language
This approach matters in environments where you:
-
Cannot trust GPS
-
Cannot rely on stable connectivity
-
Do not want humans entering first
Examples include:
-
Search and rescue in smoke, rubble, basements, or tunnels
-
Industrial inspection of pipes, corridors, and confined infrastructure
-
Defense and logistics in contested or jammed environments
From GoodAI Research to a Product
Prometheus grew out of years of GoodAI’s research into LLM agents, multi agent systems, spatial reasoning, simulation, and autonomous control. The first phase was research driven: prove that we can control real drones with agentic LLMs, perform indoor SLAM and navigation, take off and land autonomously, and self charge.
Now the bottleneck is shifting from “can we make it work?” to “how do we turn this into a reliable product and business?”
That’s why we’re spinning the project out into GoodAI Swarm Robotics. Prometheus remains the core system and brand. GoodAI Swarm Robotics will focus on productization, pilots, and commercialization.
On the product side, our current approach is:
-
Build and sell complete systems: our own drones and docks with the full Prometheus stack
-
In parallel, explore licensing the software and AI as a swarm brain that could run on other manufacturers’ drones
Over time, we want Prometheus to become both a reference hardware platform and a “swarm OS”.
Entering the Next Phase
In one sentence: GoodAI started Prometheus as an internal research project, and now we are spinning it into GoodAI Swarm Robotics to turn it into a real product and company.
The next steps for us are clear:
-
Increase robustness and reliability in real world environments
-
Move from single drones to multi drone swarms operating inside real buildings
-
Run pilot deployments with real users
-
Build the business side around this technology
And for that, we need people.
We Are Looking for a Founding CEO and Pilot Partners
To take Prometheus from prototype to operational product, we are looking for a Founding CEO. Someone who enjoys talking to customers, running pilots, raising a Seed round, and building and scaling a focused team. A hands-on operator with experience in robotics, AI, or deep tech.
We are also looking for pilot partners, especially:
-
Fire brigades and rescue teams
-
Industrial inspection and construction companies
-
Security and critical infrastructure operators
If you want to explore autonomous indoor drones and swarm capabilities in real scenarios, we would love to collaborate.
Contact
- Prometheus page: https://www.goodai.com/swarm-robotics/
- CEO / careers: https://www.goodai.com/careers/
- Direct: info@goodai.com
Leave a comment