Is FSD v14.3.3 only rolling out to EAP at this stage? If anyone from Tesla reads this and can add me to EAP that would be great. ๐
Also really happy and surprised to be on v14.3.3 - this is exciting as it opens the door for summon. Thanks @Tesla_AI and @TeslaAUNZ
Hey @davemccowen why are we fearmongering with the EV stories?
You left this nice little fact till the last paragraph, when it probably should have been the first.
"EV Firesafe director Emma Sutcliffe says there are only 14 confirmed cases of lithium-ion car battery fires in Australia, despite the nation having about 500,000 electric and plug-in hybrid vehicles on the road."
This translates to a 0.0028% chance, or roughly 1 in 35,714 vehicles.
To put this in perspective, historical data consistently shows that conventional petrol and diesel vehicles have roughly a 0.1% chance (1 in 1,000) of experiencing a fire over their lifespan, making internal combustion vehicles significantly more fire-prone than EVs.
Seems ironic that you were promoting a hybrid winning car of the year in 2024.
This is why AI safety and alignment matter.
โIf you were a hedge fund or private equity fund and you said, โWell, all I want my AI to do is maximize the value of my portfolio,โ then the AI could decide, well, the best way to do that is to short consumer stocks, go long on defense stocks, and start a war.โ
โ @elonmusk
This example was Claude 4.6 (a frontier model) and it just demonstrated the exact same tendency on a simple personal project. Speed over compliance; every time.
This isnโt just a Claude problem. In principle itโs the same for any model. The โcontractโ could be a tiny project boardโฆ or starting a war. AI Safety matters
#AISafety #FML #FACEPALM
SpaceX faced additional challenges to achieve Starshipโs full reusability. First, they had to make even non-reusable launches cheap enough to drive up cadence, which lowered prices further and unlocked more use cases. Then they introduced partial reusability to repeat the cycle. Starlink was the critical breakthroughโit created the high-volume demand and infrastructure needed to justify full reusability. Only after that could they tackle the final engineering for complete, rapid reuse.
@elonmusk Is there any reason why XAi doesn't allow subscription style use within VSCode similar to other frontier models? This would increase adoption significantly along with a native VS integration
@wholemars@DirtyTesLa The original promise was that teslas would be an appreciating asset (may be true in the future but certainly is not now) and it was implied that FSD would be unsupervised. The name changes came later.
Especially for the earlier adopters who are stuck with hardware 3 in ageing vehicles, this feels like a kick in the teeth.
Ideally anyone who's got a car more than 4 years old which would imply pretty much hardware free owners effectively retaining amnesty to transfer until either hardware solution is sorted or unsupervised is delivered with a grace period of 6 to 12 months for that transfer cuz you know people don't really have money lying around all the time. This would be a small subset of users realistically against the larger customer base. The early adopters really did help Tesla here. Otherwise other competitors (competence aside) would have succeeded as well, which they haven't at least to the same degree. For a company that for all intents and purposes does not advertise, user sentiment is critical to brand and company success.
@elonmusk I was thinking about adversarial attacks on Optimus like humanoid robots...it might just be as simple as spouse spiking food/ingredients in a kitchen when Optimus makes lunch ๐ฑ
At some point we will need a digital analogue to smell and taste, perhaps a simplified spectrometer
Excusi? @newscomauHQ WTF!
Look for and connect to "ANY" device on my Local network?
I was already barely using your site, but now I'm definately not using your sad excuse of a an adware site website anymore! Cya
#CyberSecurityRisk
O don't foresee uptake percentage on the fleet to increase with a higher price even taking account the greatly increased capability. Let me explain.
Public figures on exact subscriber numbers aren't out there right now, but based on Tesla's own Q3 2025 earnings call (still the latest cited benchmark in Jan 2026 discussions), fleet-wide FSD adoption is around 12% โ definitely well under 30%, even with higher rates in premium models or enthusiast groups.
I don't disagree that unsupervised FSD's "sleep or phone" capability represents a massive value jump, so it makes sense for Tesla to price supervised higher as it improves (and eventually charge a premium for the real unsupervised experience). But scaling to millions of subs, let alone Elon's 10M target, will be tough at $99+/month for broad adoption, especially without a true mass-market vehicle pushing the HW4+ fleet toward 20M+ units anytime soon. At that scale, you'd still need ~50% take rate to hit big numbers, which feels ambitious given current price sensitivity.
For mass-market appeal, I suspect a sweet spot closer to $50/month (or even lower) once unsupervised is live and proven โ enough to feel like a no-brainer upgrade without alienating the average owner. Early buyers/subscribers will likely come out ahead anyway, since they'll capture the biggest cut of any future robotaxi earnings potential on top of Tesla's take.