Reviewed pitch deck / rebuilt in current Senlay style
Reciprocal verification network
Senlay.
What is happening here? Can I operate here safely? How sure are we? Contribute a sensor and get the network back; consume without contributing and pay.
Problem
AI systems are entering real workflows. They need physical-world evidence.
- Agents plan drone missions, outdoor crews, vessel routes, and crop timing.
- They reason in text. They do not measure conditions.
- They rely on generic forecasts, stale dashboards, device pings, or training-data guesses.
Wind, gusts, waves, air quality, terrain, tides, and sensor freshness change the answer. A confident generic answer can be operationally wrong.
Insight
The missing layer is not more weather data. It is decision-ready physical evidence.
Senlay: "wind is 18 knots, gusts are rising, the live station is fresher than the forecast, confidence is medium-high, and this matters for this decision."
Product
One verification request. Live physical-world verification for a coordinate.
Agent asks about a place and domain.
Senlay gathers sensors, models, marine, air, terrain, and hazard signals.
Values carry source, freshness, distance, and confidence.
Live and forecast disagreement is surfaced.
Response includes agent-ready decision context and risk-event output when the workflow needs escalation.
Demo
Drone go/no-go in one verification request.
"Can a drone safely fly here, now?"
Why now
Agents are moving from chat into real-world workflows.
Chatbots
LLMs answer questions and draft text.
Copilots
LLMs help humans inside tools and workflows.
Operational agents
Agents schedule, route, plan, and trigger real actions. Physical context becomes infrastructure.
Market and wedge
Water sports is the wedge, not the ceiling.
| Stage | Market | Why it works |
|---|---|---|
| Now | SmartSurf schools, stations, and serious riders | The domain where the founder can tell instantly whether the answer is true. Proving ground for the evidence engine. |
| Now, parallel | Contributors: stations, marinas, farms, research sites | Attach a sensor, receive the network back. No cash. Coverage and distribution, not revenue. |
| Revenue | Drone/BVLOS, robotics, lifting, insurance, marine ops | Systems that act and will never install hardware. They need the answer, not the network. |
Business model
Contribute or pay.
| Mode | Price | For |
|---|---|---|
| 1 · Contribute a sensor | $0 forever | Coverage and distribution — not revenue. |
| 2 · Developer API | $0–$29/mo | Outdoor apps, IoT products, agent frameworks. Self-serve. |
| 3 · Operational safety | from $199/mo | Structured risk events for schools, stations, fleets — the layer SmartSurf validates. |
| 4–6 · Vertical / network / private instance | custom | Insurance, agriculture, drone ops; reciprocal fleet integration; bring-your-own-sources. Direction, not committed timeline. |
Revenue comes from systems that act and will never install hardware. Coverage comes from people who already own instruments. The two sides do not compete for the same budget.
Moat — named honestly
WeatherXM and Synoptic are inputs, not opponents. The moat is trust scoring, not raw data.
~10,000 stations, 81 countries, ~$13M raised. A good weather network — and a source family Senlay reasons over, not a competitor.
A known risk in volunteer networks. Senlay exposes freshness, distance and source disagreement so operators can evaluate a station instead of treating every reading as ground truth. Calibration and neighbour checks remain pilot work.
Documentation moves the starting point. Behaviour moves the weight, permanently, in both directions — not purchasable.
What compounds is not station count. Senlay can retain descriptive local history and compare it with current sources; improvement is a measured hypothesis, not a promise based on query count.
Data sources and language models are interchangeable. The judgement between them — normalization, weighting, divergence, refusal to infer — is not.
Giving network access for data attacks a vendor's own pricing. For Senlay it's the growth mechanism.
Founder
Founder-market fit through 20 years of real-world wind, water, and risk.
- Sole human founder, Hoi An, Vietnam.
- 20+ years certified kitesurfing instructor.
- IYT International Bareboat Skipper, Power & Sail up to 24 m.
- 1NCE Certified IoT Integrator.
- Built SmartSurf GPS safety tracker hardware before Senlay.
The product value is not knowing that wind exists. It is knowing what a raw number means in field conditions, where sensor placement, gusts, exposure, water state, and weak signals change the decision.
Traction and roadmap
Live platform. Open contributor protocol. Building toward external usage.
| Window | Proof |
|---|---|
| Live today | API, docs, demo, dashboard, MCP server, open contributor protocol (Apache 2.0), HMAC-signed station ingestion in production, controlled source-health status. |
| Next 30 days | Signed evidence receipts, OGC SensorThings adapter, first external API keys tracked honestly. |
| Next 12 months | Contributor stations beyond the founder's own network, one reciprocal network-provider integration, paying consumers across developer and operational-safety tiers. |
Ask
$500K pre-seed. Pre-revenue, no paying customers — said plainly.
- Signed evidence receipts (closes the output half of the provenance chain).
- OGC SensorThings adapter — turns "attach a sensor" into "attach the network you already run."
- API reliability and monitoring.
- SmartSurf risk-event proof as the operational-safety tier.
- Senior backend engineering support.
- External API keys issued to people never met, with real request volume.
- Contributor stations live beyond the founder's own network.
- One reciprocal network-provider integration.
- Paying consumers across the developer/operational-safety tiers.
New live-evidence investor brief / rebuilt into the same deck
Live evidence inside
Spot-level verification for outdoor AI apps.
The newer investor brief sharpens the thesis: Senlay is not just better answers. It is better evidence behind the answer.
Failure mode
Agents confidently hallucinate about the physical world.
A browser or agent cache can replay two-hour-old "live" conditions unless dynamic endpoints are explicitly uncached.
A forecast may be reasonable for the region but wrong at the beach, ridge, field, marina, or launch site.
Without source and disagreement, the answer sounds certain even when the evidence is weak.
Live case
Hoi An: "Can I kite right now?"
The agent gives generic advice from a forecast-like answer. It cannot inspect local sensor freshness, hardware/model divergence, gust factor, waves, or terrain exposure.
The agent receives current wind, gusts, wave state, direction, source labels, distance, confidence, and a kitesurf-specific interpretation with warnings and recheck guidance.
Same question. Different answer because the agent gets measurable context.
Forecast trust score
Senlay checks whether the forecast itself deserves trust.
| Signal | Hardware | Forecast model | Interpretation |
|---|---|---|---|
| Wind speed | Nearby station first | Model trend | Use hardware as anchor; model as comparison. |
| Gust factor | Live gust/speed ratio | Often smoothed | High gust factor lowers confidence for beginner decisions. |
| Trend | Observed change | Forecast trajectory | Agreement increases forecast trust; divergence lowers it. |
| Verdict | Evidence | Prediction | Do not hide disagreement inside one number. |
Sensor hierarchy
Closest valid hardware anchors. Farther hardware audits.
Find nearest real hardware sensor with valid QC.
Apply minimum local modifiers: terrain elevation where available; marine depth modifiers only with validated bathymetry.
Use farther sensors to compare trend and detect suspicious divergence.
Compare live observations with forecast models.
Return trust score and explain why the forecast can or cannot be trusted here.
This rule applies across locations and domains: nearest trustworthy measurement first, model and remote sensors as context.
Founder velocity
Agents found real bugs. Senlay shipped fixes and re-verified.
- Kelvin values passing through as Celsius.
- VPD unit confusion.
- Agent-facing endpoint cache risk.
- UI mode mismatch between kitesurfing and drone prompts.
Senlay is being tested by the exact class of customer it serves: AI agents that need physical context, source freshness, and clean interpretation. The product improves under agent pressure.
Thesis
Not a physical context layer. What is happening at this spot, and can something go out there.
Concrete, coordinate-level, outdoor, multi-domain, safety-relevant, and explicit about the limits of its own knowledge. Coverage comes from participants; revenue comes from systems that act.
Signed evidence receipts, OGC SensorThings adapter, reliability, monitoring.
External API keys issued to people never met, contributor stations beyond the founder's own network, first paying consumers.
Tip: use the browser print dialog to export this rebuilt HTML deck to PDF. The source combines the reviewed 12-slide pitch materials with the 7-slide live-evidence investor brief.