A Prototype Is a Beginning, Not a Deployment
Creating a GPS-denied drone in 24 hours using AI for defense applications.
Original episode title: Creating a Revolutionary New Drone in 24 Hours, w/ AI Engineer Ian Laffey
Guest
Ian Laffey
AI engineer
Ian Laffey was chief technology officer at Theseus, a defense-technology startup. He helped develop a low-cost drone designed to operate without conventional GPS signals, and has discussed the engineering, manufacturing, and AI-assisted design decisions behind rapid hardware experimentation in contested or communications-degraded environments.
Offers a vivid prototype story while exposing the difference between a fast build, a tested capability, and an operationally validated system.
What this conversation is really about
Ian Laffey’s rapid drone-navigation build is an unusually clear demonstration of what modern tools, open-source components, and a focused team can assemble quickly. The system reportedly uses camera imagery and map data to estimate location without a live GPS signal. That makes the story vivid; it does not establish battlefield reliability, safety, market readiness, or operational use. The episode is most useful as an evidence ladder: idea, hackathon build, controlled test, adversarial evaluation, field validation, and deployment are different claims.
From the conversation
The argument in focus
Ian Laffey
AI engineer
Ian Laffey was chief technology officer at Theseus, a defense-technology startup. He helped develop a low-cost drone designed to operate without conventional GPS signals, and has discussed the engineering, manufacturing, and AI-assisted design decisions behind rapid hardware experimentation in contested or communications-degraded environments.
Offers a vivid prototype story while exposing the difference between a fast build, a tested capability, and an operationally validated system.
Evidence status
Pilot / prototype
A builder reports a rapidly assembled hackathon prototype using imagery and map data for GPS-independent navigation. Speed and reported performance do not establish battlefield validation, reliability, or deployment.
Boundary map
Where the system stops
- What the system handles
- Image matching, map comparison, estimated positioning, and selected navigation support without relying on a live GPS signal.
- What remains human
- Mission definition, target and use authorization, testing, safety review, deployment decisions, dual-use judgment, and responsibility for failure.
- What remains open
- What survives controlled testing, adversarial conditions, poor imagery, changing terrain, and the legal and moral requirements of operational use?
Ideas worth carrying forward
- Report the stage of a system as carefully as its capability.
- Separate build speed from reliability under changing conditions.
- Test dual-use prototypes against adversarial and legal scenarios early.
- Do not turn reported interest or a viral demonstration into adoption.
What this changes Monday
Take one celebrated prototype and write a stage label beside every claim. List the environments it has and has not faced, the failure modes that matter, and who authorizes further testing. Define the evidence required to move from demonstration to pilot and from pilot to operational use. If those gates are missing, momentum is substituting for evaluation.
Original episode notes
At a recent hackathon, Ian Laffey was part of a team that designed, created, 3D-printed, and assembled a drone in 24 hours... and for under $500.
The drone's breakthrough: It doesn't need a signal for GPS, which means that it can be deployed off the grid, and can't be "jammed" in a war zone...like Ukraine.
Laffey's drone is now a sensation in the world of defense-tech.
Laffey is now the CTO of Theseus, and he joins the pod to discuss how his team pulled this off, why the drone is such a breakthrough, how AI played a crucial role -- we get into the nuts and bolts, not just platitudes -- and then Laffey shares, on a day-to-day basis, what it's really like to be a 24-year-old AI engineer working in San Francisco, the world's nexus for AI innovation.
Fun convo!
Ian Laffey on Twitter/X:
https://twitter.com/ilaffey2
Theseus:
https://theseus.so
Article from AviationWeek on the drone Laffey helped create:
https://aviationweek.com/aerospace/emerging-technologies/how-trio-engineers-developed-gps-denied-drone-under-500