Slow and ambiguous source loading
Terrain, buildings, vegetation and airspace could fail or arrive out of sequence without a dependable project-level truth state.
Product and engineering case study
Turning a plausible prototype into an inspectable professional beta.
SighThor began as an AI-generated idea for two hard planning problems: drone route visibility and terrain-aware camera coverage. The modernisation focused on calculation truth, explicit uncertainty, accountable review and a professional working experience.
01 / The problem
Attractive geometry is not enough when the output informs real work.
The original product could draw useful-looking cones and routes, but data failures, inconsistent terrain contexts, weak report lineage and generic navigation made the result hard to trust or explain.
Plausible but disconnected geometry
Camera, route, 3D and export paths could interpret a different scene or make hidden flat-world assumptions.
A generic dark SaaS shell
The public page advertised atmosphere while the app exposed dense options without a clear professional next action.
02 / Research and architecture
The rebuild started with boundaries, not decoration.
UK aviation requirements, terrain and LiDAR access, camera-design practice, source licensing and competitive workflows informed a phased architecture around frozen scenes and human-reviewed evidence.
- Product position
- Professional simulation and evidence acceleration — not an automatic permission service.
- Data model
- Preview, Enhanced and Verified grades remain separate from finding and report lifecycle states.
- Calculation contract
- One scene snapshot binds configuration, source evidence, engine version and deterministic outcome.
- Failure contract
- Unknown, partial, stale and failed evidence cannot silently become flat terrain or a clean result.
- Delivery contract
- Independent Sherlock gates review engineering, customer experience, claims and live acceptance.
03 / Before and after
The visible product now follows the evidence lifecycle.
The centered dark hero and passive project gallery are replaced by a paper survey brief, separate edition doors, frozen product proof, an operations register and persistent workflow readiness.
Hillcrest Survey Corridor
Five route points · two observers · 80 m AGL
Terrain sourcefixture-terrain-surface-v1
DatumSYNTHETIC-AMSL
Resolution1 m fixture
FindingBlocked + unknown
Riverside Logistics Yard
Six cameras · 60,000 m² synthetic site · nominal pixel density
Surface sourcefixture-site-surface-v1
OpticsManual FOV
Cameras6
FindingOcclusion + unknown
04 / Safety boundary
The work is strongest where it stays honest.
Wales Enhanced remains blocked pending authoritative datum evidence. Paid OSNI data remains disabled. Geometric VLOS and nominal DORI remain inputs to practitioner review, not automatic professional decisions.
Delivered capability
- Server-side bounded terrain and surface evidence.
- Deterministic camera and route analysis foundations.
- Visible source health and immutable provenance.
- Professional workflow and report-lifecycle architecture.
- Anonymous frozen proof and public methodology.
Independent work still required
- Aviation specialist review of UK rule packs and evidence wording.
- CCTV practitioner comparison with measured field scenes.
- Closed-beta feedback from operators and consultants.
- Source operations and licensing for each enabled deployment.
- Expansion only after credible demand and validation.
Consulting capability
Geospatial products, evidence workflows and professional product modernisation.
This work demonstrates research, domain modelling, data provenance, deterministic analysis, security, workflow design, report architecture, testing and deployment verification. A consultation can explore a similar product or a private deployment requirement without implying regulatory authority.