EXCAR isn't a rendering or a pitch deck. It's running on real hardware in a real car today, listening and talking back. Here's exactly where it stands, and what's left to build.
One is what exists. The rest are the gates we're building toward, in order.
The full pipeline of listening, thinking and talking back runs end to end on the prototype hardware right now.
We'd rather show you what has to be true to call a stage done than quote a completion percentage nobody can check. Pick a stage.
Wake and push-to-talk conversation, streaming speech, music by voice, navigation handed to the phone, and memory that survives a reboot. All of it working together in a vehicle that actually drives.
Everything after this depends on one question: does the loop close in a moving car, with road noise, a real driver and no laboratory? It does. That answer is the foundation the rest of this list is built on.
Measurement, mechanical design, electrical and power behaviour, thermal limits, cabin acoustics, the Bluetooth OBD link, and testing in the vehicle rather than on a desk.
A car interior is a genuinely hostile place for electronics: heat, vibration, noise, and a battery with other priorities. Every unknown we close here is one we don't pay for later, in someone else's car, at ten times the cost.
The prototype is large on purpose. Development boards and off-the-shelf parts let us prove the idea in months instead of years, and they made it bulky. A custom board is how that becomes a product.
Nobody wants a workbench on their dashboard. Our own board is what turns a proof into something smaller, quieter, cheaper to build, and designed for a car interior instead of a desk. It is also the first point at which unit economics become real.
Small enough that we fit every unit ourselves and speak to every driver. Large enough that real-world variation starts telling the truth.
A demo tells you the product works. Ten cars over weeks tell you whether people still use it in week three, which is the only usage number that means anything.
The prototype's answers came from a cloud API. That proved the product, but the character at the centre of it was rented. EXCAR LLM, trained from open weights, is how it becomes ours.
A companion people form an attachment to cannot be a thin layer over someone else's chatbot. Owning the model is how EXCAR's judgment, memory and voice stay consistent, and how latency and cost stop being another company's decision.
A wider spread of makes, model years, cabins and owners. This is the point at which our claims meet hardware variation we didn't choose.
Ten cars tell you whether it works. Fifty tell you whether it works without a founder in the passenger seat. That difference is the whole business.
Beyond stage 05 is opening EXCAR to everyone who wants one. We're not putting a date on that, because the honest answer is that it depends on how the fifty cars behave.
What we learn from real drivers defines the board. What the board can carry defines the intelligence it serves. Neither track finishes alone.
Smaller hardware. Deeper relationship.
EXCAR's software is split into a car-agnostic core and small per-vehicle plugins, so supporting a new car model is an addition, not a rewrite.
Wake word detection, speech-to-text, the conversation engine, text-to-speech, and memory, identical in every car, regardless of make or model.
Per-vehicle integration lives in a small plugin, so supporting a new make or model is an addition, not a rewrite.
Every response runs through a cloud language model tuned for this environment. We're training our own, EXCAR LLM, from open weights, so the character behind every conversation is ours.
Join the waitlist for the first pilot, or follow along as EXCAR moves from one car to a fleet.