How we went from no anti-theft to a network-scale, lawful stolen-asset recovery grid — in about a week, built by a team of AI agents on a transport that was ready for it from day one. This is less a feature and more how we build features.
จากศูนย์สู่ระบบกู้คืนทรัพย์สินที่ถูกกฎหมายบนโครงข่ายเมช — สร้างในเวลาประมาณหนึ่งสัปดาห์
ทรัพย์สินที่ถูกขโมยมักหายไปในจุดที่สัญญาณมือถือเข้าไม่ถึง — เราครอบคลุมจุดเหล่านั้นอยู่แล้ว
Mall basements & podium parking — cellular-dead, mesh-reachable.
Garment-market lanes like Pratunam — congested cell, high theft exposure.
Rural & disaster zones with no carrier backhaul at all.
[EXT — cite], and today's
detection is expensive and short-range. Partnership pending.เมชคลื่นต่ำครอบคลุมพื้นที่กว้างในราคาเพียงเศษเสี้ยว และตรวจจับสถานีฐานปลอมได้ทั้งเครือข่าย
Concrete-penetrating, off-grid, no carrier backhaul required.
~$150 gateway + ~$30 nodes vs telco capex for the same footprint.
A grid sees what one phone can't → the partnership the telcos need.
Anti-theft is our first public-good offering to a societal problem — cheap to build because the mesh nodes are the location (RF-attested, not spoofable GPS) and the wire protocol reserved frame types from day one. Payloads on a proven transport, not pivots.
Shipped on it already · built + CI-tested
Sighting spine, reverse-EIR POS gate, and crowd-BLE — all shipped & CI-tested (detailed on the next slides). Native always-on BLE source is pending hardware.
Deep-indoor current-draw over the mesh, live. The on-node current sensor on real hardware is next build.
Network-scale fake-base-station correlation engine, CI-tested. On-node modem sensor pending partner; network-wide rollout planned.
Reserved & ready to activate · designed-in headroom, not yet built
Missing-person / SAR on the reserved type-4 priority frame + the dual-signature override primitive. Workflow not yet built.
Herd & high-value-asset tracking in the cellular blindspots farms live in. Planned — no code yet.
PDPA-clean aggregate telemetry — rainfall, river level, crowd density (frame types 5–7 reserved).
ที่ส่งมอบแล้ว — สร้างและทดสอบจริง · ส่วนที่สงวนไว้พร้อมเปิดใช้งานแต่ยังไม่ได้สร้าง
Each step is one payload on the same spine, logged and CI-green. Version tags are our own changelog (alpha.N).
A commerce-only mesh: payments & dispatch.
Report-stolen → node hears it → RF-attested position → owner alert. Interception rejected.
A grid correlates fake-base-station signatures — the partnership anchor.
Deep-indoor coverage story, on a live 100-node map.
A second radio on the same spine.
Highest-risk radio built inert; the honest decision-of-record.
Every radio plugs into one network-agnostic pipeline. Adding a new
way to hear a stolen asset is a new source tag — not new tracking machinery.
Owner adds an asset + serial/IMEI.
Token activated (signed, audited).
Mesh, BLE, POS scan, carrier…
RSSI-weighted from the nodes that heard it.
SMS + LINE, with a map.
| Source | How it hears | Status |
|---|---|---|
mesh | LoRa nodes hear the tag beacon | built · CI-tested |
ble | Merchant terminals hear a registered recovery tag | built native always-on: pending hardware |
pos-gate | A scanned serial matches a stolen tag at checkout | built · CI-tested |
carrier | Operator's consented CAMARA device-location API | awaiting operator agreement |
helium / meshtastic | Adjacent public LoRaWAN networks | awaiting credentials |
การเป็นเจ้าของโครงข่ายเองคือความได้เปรียบที่คู่แข่งที่พึ่งสัญญาณมือถือลอกเลียนไม่ได้
Because they all sit on one spine, each is a module we can turn on per deployment. Status shown honestly.
A scanned serial matching a stolen tag fails the sale closed. A lawful pawnshop-style interlock, cross-tenant like a national stolen registry.
Merchant terminals hear owner-registered recovery tags → the sighting spine. Native always-on source is pending hardware.
Per-zone current-draw over the mesh, live — including zones with no cell signal (the coverage badge).
Built but off by default (inert) — hash-at-ingress only; awaiting Thai legal sign-off before it can run.
A live deployment plan over the real Pratunam district with asset-tags + no-cell overlays. Try the interactive demo →
Report-stolen → RF-attested position → owner alert, proven end-to-end in CI (TestRecoveryLoopEndToEnd).
We already have our own network planner. So the demo isn't a slide — it's the real map, deploying node by node.
Speed without discipline is just risk. Three things keep the build honest and de-risked end to end.
A documented roadmap and whitepaper are the throughline every build follows — the "red thread." Contents available to investors under NDA.
Every feature carries a [subject to Thai counsel] review by a real person on the law team before it clears. Our honest labels are that gate, made visible.
The pilot scaled down from a heavy initial compute requirement to a small concrete ask — validated on free rented hardware (FIT IoT-LAB, exp. 443862 Saclay); the measured capture rate re-seeded our simulator. The pilot BOM is the RFQ (~$575–950/cell), not a guess.
| Step | What | Status |
|---|---|---|
| Done | Sighting spine + 6 sources; POS gate; energy dashboard; recovery loop proven in CI | built · CI-tested |
| Now | Native always-on BLE source on node hardware; cable bench for the 920 MHz driver | pending hardware |
| Pilot | Bangkok / Udon deployment on frozen, terrain-validated placement; carrier + Helium credentials | partner-gated |
| Scale | ~2000-node grid, rogue-cell partnership, Wi-Fi only on written counsel sign-off | counsel + partner |
We built a lawful, network-scale stolen-asset recovery grid in about a week — because we own the transport, reserved room for it from day one, and build with agent teams that ship the seam and label the metal. Anti-theft is the first payload. The pipeline is deep.
โครงข่ายเดียว หลายประโยชน์สาธารณะ — เริ่มจากระบบกันขโมย
docs/LAWFUL-RECOVERY.md, or the
main investor deck.