The products / Navigatek 6Ai
01 — Navigatek 6Ai
The command post of the six AIs
Six specialised AIs work together during the race: wind, tactics, race direction, analysis, sensor watch and literature watch. 6Ai is where you see them work, where you talk to them, and where you adjust what they do.
- 0.1 smedian, from deciding to speaking
- 84.5 %correct direction on the wind ahead of the bow, out of sample
- 250+settings adjustable at the dock, each one documented
Free with the phone's GPS alone (soon); licensed with the sensor kit or your instruments.
Why
On the water, nobody has time to read a screen.
Racing instruments display numbers. But the helm is watching the sails, the puff coming in and the boat alongside — not a screen. And a number does not say what to do with it: a polar percentage does not tell you whether to tack.
Real coaching is given by a coach in a RIB, to a few crews, a few days a season. The rest of the time you race alone, and debrief from memory.
Without
Numbers on a screen that have to be read, interpreted and turned into a decision — while steering.
With Navigatek
An instruction spoken aloud, with its reason, at the moment it matters. Hands stay on the helm, eyes on the water.
What
A crew of six AIs, and a command post to watch them work.
On the water the helmsman has his hands and his ears, and nothing else. Coaching comes out of the boat's own speaker, in plain language, while he steers — nothing is ever asked of him. 6Ai is what you open before leaving the dock and what you read on the way home: it is where the system accounts for itself.
In service on a J/70 since summer 2026: voice coaching, VMG and polars, manoeuvre analysis, post-race report. The proof, numbers included, is in the sample report.
Live
Coaching speaks during the race, through the boat's speaker. Three seconds to act on an order, and the reason comes with it: “the other tack gains on the mark, 3.9 knots better”.
Visible
Every exchange between the AIs is logged and readable. You are not asked to trust a black box: you are shown who spoke to whom, when, and what was decided.
Adjustable
No threshold is frozen in the code. Each one is adjustable at the dock, documented, and carries the measured number that justifies it.
Honest
When a reading becomes doubtful the system downgrades itself and says so, rather than serving something false. An incoherent compass falls back to GPS heading, and the crew is told.
The versions
Start with a phone, go all the way to the boat's sensors.
No installation is required to start. Each step adds measurement, and therefore advice.
- 6Ai Mobile
Free · soon
The phone's GPS alone, no installation. To begin with, what Navigatek Bord already does: spoken countdown, line alarm, speed, heading and provisional result — linked to Juge, or used on its own. Then, everything 6Ai can draw from GPS alone.
- 6Ai Kit
Licensed
The kit of everyday onboard sensors — two GPS units, wind, compass, barometer, about $1,000 — and all six AIs: live voice coaching, wind map of your water, post-race report.
- 6Ai Integration
Licensed · planned
Your existing instruments, read on the boat's network (NMEA 2000), checked like any other sensor — on the roadmap.
How
From sensor to sentence, in four steps.
- Measure
GPS at 10 Hz, ultrasonic anemometer, inertial unit, barometer. Every reading is timestamped and checked against the others: a sensor that drifts is downgraded, not believed.
- Compute
True wind, target angle and target VMG from the boat's real polar, laylines, headers. All of it deterministic: the same situation gives the same answer.
- Decide
The specialised AIs weigh their readings — the route, the water, the header, the fleet — and a single order comes out, naming the capability that asked for it and its reason.
- Speak
The language model puts the decision into a sentence a helm can act on in three seconds, and the speaker says it. Back at the dock, everything can be reread and questioned.
What's new
What 6Ai does that no instrument does.
Four design choices. Each was born of a fault met on the water, and each is measured.
- The AI words it; it does not decide
Tactics are computed by deterministic tools, from the sensors. The language model only comes in at the end, to speak. No tactical hallucination — and a whole race of coaching costs less than the report that follows it.
≈ 3 times fewer tokens than writing the report
- The water, learned square by square
The system learns how the terrain bends the breeze, in hundred-metre squares, outing after outing. It says “dead ahead, it lifts” before the boat gets there.
84.5 % correct, on outings it had never seen
- A system that knows when it is wrong
Every sensor is watched by the others. An incoherent compass downgrades itself to GPS heading, and the crew hears about it. A queued order is re-checked as it leaves the speaker: what was true twenty seconds ago may no longer be.
Degrade rather than serve something false
- Independent of any model
No AI knows which model it is talking to: it asks for a role, and a router chooses. A new model earns its place on a test bench, not because it is new. And everything can run without the cloud.
10 tasks · 50 checks · passed 3 times in a row
This is the piece that makes the other five visible. Without it, six AIs working well look exactly like six AIs doing nothing.