Liss argues that CarPlay simply streams an H.264-encoded UI rendered by the phone to a display the car already owns, with touch/knob inputs going back. The car's actual software — ECU, CAN bus, HVAC, BMS, ADAS — never sees the iPhone, so framing this as Apple 'taking over' the vehicle is factually wrong.
By submitting Liss's post and driving it to 327 points, sprawl_ amplified the position that the trade press keeps mischaracterizing CarPlay 2. The HN community's strong engagement signals broad agreement that the 'takeover' framing misrepresents what is fundamentally a compositor layer.
The editorial argues reporters translate 'GM wants the $135/month connected-services revenue for itself' into 'Apple is trying to take over your car' because the commercial story is boring and the technical one is inconvenient. The rejection of CarPlay 2 is about subscription revenue capture, not defending vehicle operating systems from Apple.
On July 2, Casey Liss published a short, exasperated post titled *CarPlay is Additive, You Dolts*. It hit 327 on Hacker News within a day. The trigger was another round of trade-press hand-wringing about GM, Rivian, and a handful of European OEMs 'rejecting' CarPlay — framed, as usual, as automakers heroically defending their operating systems from an Apple takeover.
Liss's argument is one paragraph of actual content stretched over a lot of italics: CarPlay is a projection protocol. Your phone renders a UI, encodes it as H.264, and streams it to a display the car already owns. Touch and knob inputs go back the other way. That's it. The car's actual software — the ECU firmware, the CAN bus arbitration, the HVAC controller, the battery management system, the ADAS stack — never sees an iPhone at all.
CarPlay 2, the version GM and others actually object to, extends the projection to instrument cluster and climate screens. It still doesn't touch the underlying vehicle software. It reads sensor values over a defined API and draws prettier pixels on top. If you unplug the phone, the car works exactly as it did before, because nothing was ever replaced.
The reason this keeps getting reported wrong is that the interesting story isn't technical, and reporters covering the auto beat don't want to write the boring commercial one. So they translate 'GM wants the $135/month connected-services revenue for itself' into 'Apple is trying to take over your car,' which reads better and is completely false.
The technical reality has been public since the CarPlay 2 announcement in 2022 and hasn't changed. Apple ships a spec. The OEM implements a small daemon on the head unit that speaks the wireless CarPlay protocol, exposes a defined set of vehicle-signal reads (speed, fuel, climate zones), and grants the phone permission to render into specific display regions. Everything Apple 'controls' is a compositor layer. The safety-critical software — the stuff that has to pass ISO 26262 functional safety — is untouched, because Apple has no interest in taking on liability for a brake-by-wire bug.
Compare this to what a real takeover would look like: Android Automotive OS, which actually is the operating system on Polestar, Volvo, and the new Renaults. Google's code runs the head unit. The OEM licenses Automotive Services (Maps, Play Store, Assistant) on top. That's a genuine software substitution, and notably, the same trade press that hyperventilates about CarPlay barely mentions it, because Google didn't put out a keynote video with cinematic gauge cluster b-roll.
The community reaction on the HN thread mostly agreed with Liss but drilled into the actual mechanism. Several ex-automotive engineers pointed out the piece Liss left implicit: OEMs aren't defending a technical moat, they're defending a P&L. The infotainment stack is the only part of a modern car with a plausible recurring-revenue story. Heated-seat subscriptions, remote-start apps, over-the-air feature unlocks, telematics data resale — all of it lives in the layer CarPlay covers with a phone screen. Every minute a driver spends in CarPlay is a minute they aren't looking at the OEM's music app, its navigation ads, or the upsell for the extended trial of SuperCruise.
GM said the quiet part out loud in 2023 when it announced it was dropping CarPlay from its EVs to build its own Ultifi platform. The pitch to Wall Street was $20-25 billion in annual subscription revenue by 2030. The pitch to drivers was 'a more integrated experience,' which they promptly demonstrated by shipping a Blazer EV with an infotainment system so broken GM had to stop-sale the car for a month.
If you build anything in or around automotive — head-unit apps, fleet telematics, or the OBD-II ecosystem — the framing matters, because it determines what you build against. Treat CarPlay and Android Auto as projection layers on a display you don't own, not as platforms you can target directly. There is no CarPlay SDK for arbitrary apps; Apple whitelists categories (audio, messaging, nav-with-a-partnership, EV charging, parking, and a few others). If your product is a driver-facing app, you're building an iOS app that opts into a CarPlay scene, not a car app.
For anyone in embedded or safety-critical work, the distinction Liss is trying to hammer in is the one that determines regulatory scope. A projection protocol lives entirely in the QM (quality-managed) domain of the head unit. The moment you cross into anything ASIL-rated, you're in a different world with different vendors, different toolchains (typically AUTOSAR, sometimes QNX or Integrity underneath), and a validation cycle measured in years. The 'Apple is taking over your car' framing collapses that distinction and makes it harder to have a coherent conversation about where the actual software risks in modern vehicles are — which is mostly in OEM-written code that ships with far less rigor than the powertrain firmware next to it.
For developers watching the OEM subscription push: the interesting bet is which side blinks first. Ford already reversed course and re-embraced CarPlay after Farley concluded customers would pick the phone every time. GM is doubling down. Rivian is threading the needle by shipping neither and hoping its own UI is good enough — a bet that only works if you're small enough that the phone-integration gap doesn't cost you the sale.
The boring version of this story — an OS vendor shipping a well-scoped projection protocol, and OEMs fighting a commercial battle they'll dress up as a technical one — is going to repeat every product cycle until either automakers accept that infotainment is a phone accessory or one of them ships a subscription platform good enough that drivers stop reaching for the cable. Neither side has an obvious win. What Liss got right, and what the trade press keeps getting wrong, is that it was never a technical argument in the first place.
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.