$35K airframe.
$899 brain.
The American Shahed.

1 · Why Build an American Shahed

Russia has fired thousands of Shahed-136s at Ukraine since 2022. The hard strategic lesson: cost-discipline beats cost-effectiveness. A $35K delta-wing-pusher achieves strategic effects when launched in salvos against critical infrastructure. Defending against them with $4M Patriot rounds bankrupts the defender first.

The DoD's Replicator initiative (Hicks memo, August 2023) explicitly targets "all-domain attritable autonomous systems" at the Shahed price-point. AEGIS-LM is the Switchblade-600 successor: same airframe envelope, same mission profile, but with the AI compute that Iran couldn't have because Iran can't legally buy NVIDIA silicon.

2 · Iran's Stack vs Ours

The Hackaday teardown of a captured Shahed-136 air-data computer (Hackaday 2024-07-15) revealed the entire compute stack is hobby-grade. We replace it with a real edge-AI compute module while keeping every other component cost-comparable.

3 · Same Airframe, Cheaper Production

The Iranian airframe design is good. We don't need to redesign it — we need to build it cheaper, faster, and with US-compliant supply chain. The Shahed delta-wing pusher is well-suited to mass production:

The US production line is the discipline problem, not the engineering problem. Iran builds Shaheds at ~$35K because they accept hobbyist-grade components and high failure rates. We can match that price with US-fabbed PCBs (Sierra Circuits, NDAA-compliant) by using the AEGIS-LM compute module to replace expensive guidance systems — and gain capabilities Iran can't field.

4 · The Software Stack — What the Jetson Buys You

The compute upgrade isn't bragging rights — it's mission capability that makes AEGIS-LM strictly better than the Iranian original at every operational pressure point. Five concrete capabilities:

4.1 GPS-Denied Autonomous Navigation

The Iranian Shahed flies preprogrammed waypoints under GPS+GLONASS. Russia's Kometa-M CRPA reduces but doesn't eliminate jamming vulnerability. Drop GPS and the Iranian airframe becomes useless within minutes of deviation.

AEGIS-LM runs VINS-Fusion (visual-inertial SLAM) on the Jetson — fuses 6-camera optical flow with 250Hz IMU + barometer to maintain position estimate indefinitely without GPS. Reference: HKUST-Aerial-Robotics/VINS-Fusion.

# Jetson onboard pipeline (simplified)
camera × 6 ──→ feature extraction (ORB or DL keypoints, ~3ms)
              │
              ▼
            VINS-Fusion fuses with IMU @ 200Hz
              │
              ▼
            6-DoF pose estimate (drift < 0.1% over 100km cruise)
              │
              ▼
            EKF2 fuses with optional GPS · MAVLink to FC

# When GPS drops:
  - INS + VIO maintains pose for > 4 hours
  - Loop closure on landmarks corrects drift periodically
  - Terminal phase uses EO/IR target recognition (no GPS needed)

4.2 Vision-Based Target Selection

The Shahed-136 dives on a GPS coordinate. If the target moved, it misses. If the GPS coordinate is wrong, it hits the wrong building. AEGIS-LM does terminal-phase visual targeting using a YOLO11n detector trained on the target class:

# aegis-lm/perception/terminal_seeker.cpp
- 6-camera spherical view processed at 240Hz
- YOLO11n classifier trained per mission profile:
    * Maritime: ship silhouettes, port infrastructure
    * SEAD: SAM radar dishes, AAA emplacements
    * C2: communications towers, TOC tents
- Operator pre-loads target class + waypoint
- Final 30 seconds: drone autonomously refines aimpoint within
  ~5m radius of waypoint to hit highest-confidence match
- Result: hits the actual target, not the GPS coordinate

4.3 In-Flight Mission Re-Task

Shaheds are launched and forgotten. AEGIS-LM has a Doodle Labs LTE radio ($380 BoM addition) for low-rate command bandwidth — operators can re-target, abort, or reassign drones during cruise. Critical for post-launch intel: "target moved 200m east" or "scrub TGT-A2, retarget TGT-A3."

4.4 Swarm Coordination

Iranian salvos arrive in time-staggered waves but can't coordinate mid-flight. AEGIS-LM uses ROS 2 mesh networking over the LTE+ELRS radios: each drone broadcasts its position + intent, the swarm autonomously deconflicts flight paths and assigns targets dynamically.

• Auction-based target assignment — drones bid for targets based on fuel remaining, proximity, and current detection confidence
• Mutual SEAD — first drone in distracts air defense, others penetrate on alternate vectors
• Real-time BDA sharing — first impact streams imagery, follow-on drones select different aimpoints if first looked sufficient

4.5 EW Resilience

Russian Shaheds with Kometa-M CRPA still drift hundreds of meters under directed EW. AEGIS-LM's INS+VIO+Vision triple-redundancy means no single jammer or spoofer defeats it — you'd need to physically blind the cameras, jam GPS, and disable INS simultaneously. That's a $10M+ EW system per kill on a $32K target. The math inverts.

5 · Manufacturing — Iran's Production Discipline

Iran builds 6,000+ Shaheds/year at Russian Alabuga production lines. The lessons they teach about cost-discipline are the lessons US prime contractors keep failing:

1. Single-source the BoM hard. Iran uses STM32G071 because it's $2 in volume. Don't qualify three options for every part.
2. Accept failure rates. 5-10% of Shaheds fail on launch. That's still cheaper than gold-plating reliability. Build 100, expect 90 to fly.
3. Decentralize assembly. Multiple smaller lines, not one big factory. Hardens supply against single-point disruption.
4. Software updates over the airframe. Iran can't change the STM32 firmware after launch. We can OTA-update the Jetson stack via LTE — capability evolves after fielding.

AEGIS-LM target manufacturing footprint: 3 distributed lines (TX, OH, AL), each producing 200-400 units/year, with a shared US-fabbed PCBA partner (Sierra Circuits + MacroFab). Total nameplate: ~1000/year at LRIP, scalable to 5000/year under Replicator.

6 · American Design Language

The Iranian original is grey, utilitarian, and visually anonymous — that's by design (cheap to paint, hard to ID against haze). AEGIS-LM rejects that aesthetic in favor of an explicitly American military-industrial visual language. The drone is meant to be photographed at delivery, in test, and on the rail: it has to look like an instrument of US deterrence, not a copy.

6.1 Materials — DFARS-compliant US sourcing

Iran sources from grey-market Chinese carbon, Turkish servos, and Russian-routed microcontrollers. AEGIS-LM is built on a fully DFARS § 252.225-7009 compliant specialty-metals chain — every load-bearing structure traces to a US melt and US fabrication. This is not just compliance theater; it's what the contract demands and what unlocks Replicator-tier production volume.

6.2 Visual — Anduril-tactical, not Iranian-grey

The aesthetic specification is documented and enforced across all renders, demos, and physical prototypes:

• Primary skin: matte carbon-black (Pantone 419 C / Hex #1a1d22) — the same low-IR tactical finish you see on Anduril Roadrunner and Kratos XQ-58. Reads as "American military-industrial," not "Soviet-era utilitarian."
• Secondary accent: olive drab (Hex #3a4630) — applied to the warhead bay band, signaling lethality without billboard markings. Inherited visual line from M2 Bradley, M1A2 Abrams, and the Patriot launch-rail livery.
• Tertiary highlight: AEGIS cyan (Hex #00c2ad) — emissive accent at the propeller hub. This is our brand color. It's the same teal you see on Petronas F1 (Mercedes), and it telegraphs "this is a software-defined weapon" the way carbon-fiber telegraphs "this is the modern fighter."
• National marking: 5-point USAF roundel with white star on navy (#1c3050) field — wing-top placement, no callouts. Subdued low-vis variant for theater operations.
• Tail callsign: AEGIS-LM-1 in JetBrains Mono at 14pt height, snow-white on tail. Matches the deployment HUD typography.
• No "Geran-2" cyrillic, no Iranian crest, no anonymous grey. This is the point of the program: the airframe is the same shape but the visual language is unmistakably American.

6.3 Why this matters for procurement

Defense procurement is a visual industry. SECDEF photo-ops, contractor day at Fort Liberty, Replicator demos, RIMPAC press releases — they all need an image. A grey copy of an Iranian drone in those frames sends the wrong message. A matte-black AEGIS-LM-1 with a US star on the wing and Jetson silicon visible through the inspection port sends the right one: "we built our own, it's better, and it's photogenic enough to put on the cover of Aviation Week."

The aesthetic also derisks Congressional optics. Members will see this drone on briefing slides. It needs to look unmistakably American the moment it shows up on a SCIF screen — no time for staffers to wonder whether the photo is a captured Iranian drone or a US weapon.

7 · Cost-Exchange Math

For the cost of one Tomahawk ($2M), we field ~60 AEGIS-LM ($32K each). For the cost of one Switchblade 600 ($170K), we field ~5 AEGIS-LM. The Replicator math is identical to Iran's strategic insight: small, smart, cheap, and many.

8 · Mission Profiles

The same airframe runs five distinct mission profiles by software loadout:

9 · Acquisition Pathway

10 · The Cinematic — See It Fly

The accompanying cinematic preview at /r3f-strike-mission shows the full mission profile in 70 seconds: 5-unit salvo launch, climb-out, long-range cruise, terminal dive, and impact — with the hero drone always on screen.