F_nel retweetou
F_nel
564 posts
F_nel retweetou
F_nel retweetou
We
– design the chips & hardware
– make the cars w/ said hardware
– collect real-world data at scale
– train the real-world AI model
– built (& continue to expand) the massive supercomputer cluster that trains it
– deploy AI directly to millions of robots on wheels
All that is shared with @Tesla_Optimus for broader applications in both the physical & digital world
English
F_nel retweetou
F_nel retweetou
NEWS: NASA is planning a bigger @SpaceX Moon mission role using Starship, in a massive blow to Boeing.
With the new proposal, Boeing's SLS would no longer be used to boost Orion close to the moon. Instead, Starship and Orion would dock in Earth orbit, giving Starship the pivotal role of propelling the capsule to the moon’s orbit, before taking astronauts down to the surface. bloomberg.com/news/articles/…
English
F_nel retweetou
@SawyerMerritt @SpaceX SpaceX will ultimately deliver millions of tons to the Moon to build a self-growing city there and same for Mars
English
F_nel retweetou
NEWS: The NHTSA has announced that its has upgraded the probe into Tesla's FSD (Supervised) in low-visibility conditions to what’s known as an “engineering analysis.” It’s a step that is often required before the agency tells a company to issue a OTA recall, but does not automatically mean that the NHTSA will issue a OTA recall.
The NHTSA said its engineering analysis follows an earlier preliminary review and broadens the probe to about 3.2 million Tesla vehicles across multiple models equipped with the system, covering most vehicles on U.S. roads.
English
F_nel retweetou
@mark_k @cursor_ai Coding will be generically available from many companies in a few months
English
F_nel retweetou
F_nel retweetou
F_nel retweetou
We’ve been able to generate physics-accurate, real-time video for self-driving training & testing at @Tesla_AI for a long time.
The compute required for this (roughly one H100 per HD camera) is still far too expensive for consumer use, but probably becomes affordable in 2 to 3 years.
English
F_nel retweetou
@teslaownersSV @pbeisel With some luck and acceleration using AI, we might be able to tape out AI6 in December
English
F_nel retweetou
F_nel retweetou
@bindureddy Google will win the AI race in the West, China on Earth and SpaceX in space
English
F_nel retweetou
F_nel retweetou
F_nel retweetou
NEWS: Tesla has started hiring for its Terafab (chip manufacturing) project in Austin, Texas!
Job listing: "You'll own end-to-end program scoping—including factory design/construction from concept through execution, ramp-up, and production readiness. The ideal candidate brings a strong background in technical program management, semiconductor facility engineering, multidisciplinary design leadership, value engineering, capacity modeling, and cross-functional execution for large-scale programs."
Ryan Rogue@TeslaRyanRogue
Terafab JOB Posting - It’s happening. @SawyerMerritt
English
@FredLambert If, based on the point when FSD was disengaged due to intervention, the vehicle could have comfortably navigated the corner with just 2–3 seconds of deceleration, then one cannot blame FSD’s speed.
English
@FredLambert If FSD was disengaged due to driver intervention at the very moment it was trying to reduce speed, unfortunately, it means the driver failed to properly control the situation after the intervention.
English
Tesla fans using the “4-second disengagement” as a gotcha are missing the forest for the trees. Yes, the driver was technically in control of the vehicle at the moment of impact. But she was in control because FSD was already failing by driving too fast ahead of this sharp turn — it was heading straight into a concrete barrier at highway speed with no sign of correcting.
Everyone who has frequently used FSD or Autopilot and paints this 4-second disengagement as a “gotcha” moment is being disengenous, and that includes Elon Musk.
I have tens of thousands of miles on FSD, and I’ve experienced the system coming too fast into a turn at least half a dozen times.
We’ve said this before and we’ll keep saying it: the problem with FSD isn’t what happens when the driver is paying attention and the system works. The problem is what happens when the system gives you every reason to trust it, and then suddenly doesn’t work. The driver has to recognize the failure, assess the situation, decide on a correction, and physically execute it, all in less time than the system needs to create the danger.
Musk and Tesla’s propagandists can point to the logs all they want. The video shows what actually matters: FSD approaching a standard highway curve at full speed with zero indication it was going to navigate it. That’s the failure. Everything that happened after, including the panicked disengagement, is a consequence of that failure.
The framing that this was “manual driving, not FSD” is technically true for the final 4 seconds and deeply dishonest about the full sequence of events. It’s exactly the kind of liability shell game that courts are increasingly rejecting, as that $243 million verdict makes clear. Tesla created the system, sold it as “Full Self-Driving,” and profits from the ambiguity. At some point, it has to own the consequences.
Electrek.co@ElectrekCo
Tesla says FSD was off before Cybertruck crash — but the video tells a different story electrek.co/2026/03/18/tes… by @fredlambert
English
@PcIOvebbCbTdSTb 4.2 이전엔 다른 것들이랑 병행으로 썼는데 이젠 그냥 얘만 씁니다ㅋㅋㅋ 저는 그록이 감당하기 어려운 일을 주지도 않아서 지금 환각 줄어든 것만으로도 사용성이 엄청 좋아졌어요
한국어
벤치마크에서 체감 성능을 느끼는 유일한 지표는 이제 할루시네이션인듯 이전부터 클로드류가 유달리 높은 편이었고 이번에 4.20도 확실히 개선이 느껴짐
Emergence@PcIOvebbCbTdSTb
Truth Seeker
한국어