Huge news from NHTSA today! They’re launching the ASCEND consortium, a cooperation agreement with SAE to expedite the development of data-driven AV performance standards. Their top priority is to inform and accelerate corresponding FMVSS standards. “Many of the regulations were designed when no one envisioned a vehicle without a human driver, and therefore require or reference manual controls.” “This means that the technology of the future is often tethered to the regulations of the past. So we are working diligently on much-needed FMVSS updates to account for new designs enabled by automated driving systems, without compromising safety.” Their ADS performance rulemaking will provide federal leadership by delivering an AV competency standard that will hopefully address inconsistent state and local requirements. Vehicles like the Tesla Cybercab are being held back by old laws and standards that no longer make sense, and NHTSA is well aware. It’s great to see them taking this initiative quickly.

NHTSA has asked Tesla for a Special Order after the Cybercab Audit Query on September 3rd. Tesla’s response to this Special Order must be provided to NHTSA’s Office of the Chief Counsel by September 30, 2026. Tesla’s response must be signed under oath. "On September 3, 2026, NHTSA opened 26002 to evaluate Tesla’s basis for certifying that its Cybercab vehicle complies with all applicable FMVSS. This Special Order requires information related to the process and technical data on which Tesla relied when certifying the Cybercab and related issues. NHTSA is asking 21 requests questions to be answered. Below is a link to all of these questions. I personally think NHTSA will accept Teslas answers to why it does meet the FMVSS standards. The Cybercab was never meant to have driving controls as it was built from the ground up to be autonomous. I also beleive Tesla will use their safety data as a reason why they beleive the vehicle does not need these traditional controls and mirrors.

After today I will no longer support or endorse Waymo specifically. This will likely result in losing followers and/or ruffling some feathers, but so be it. I don’t care. The good news is I won’t ever need a ride from them in the future, there will be so many more Teslas, including my personal car. People say we should support all autonomy. Well, this company clearly doesn’t and is obviously threatened by Tesla, but instead of welcoming them they’re attacking them. The reason is their personal attacks on Tesla. This is not the first time they have openly criticized them, going so far as to say their approach is unsafe and cannot work and “can’t see at night”. From CEO to top engineers “you have to be able to see at night” we’ve heard it all. In the thread today they blatantly called Tesla a fraud, even though they are far too afraid to mention them by name. They of course start with: “Multimodal sensors are indispensable. By combining inputs from cameras, lidar, and radar, the Waymo Driver creates a rich, redundant world view that no single sensor can replicate.” “No single sensor can replicate…” The first direct attack against Tesla. Funny enough, Tesla has already proven this wrong with Robotaxi operating safely. As Ashok has said many times, AI is not a sensor suite problem; it’s an AI problem. “Simply improving a driver assistance system for full autonomy is a FALSE summit” This directly refers to Tesla FSD supervised and how Tesla has used this 13B miles of data to create real-world AI and Robotaxi. “You can’t build trust with a black box. AI is powerful, but it isn’t magic. Pure end-to-end architectures risk black box failures where decision-making is hard to understand.” This again directly attacks Tesla, which uses a pure end-to-end approach. Waymo claims this “risks” failures where decisions are hard to understand. Actually quite hilarious—V14 stops for school buses, yet why is Waymo investigated by NHTSA for blowing past school buses with stop signs out? 🤔 What will they say when Tesla has 10x their fleet, 100x? I can’t wait for Tesla to prove the doubters wrong and mop the floor with them. I still support the push for all autonomy, but I don’t support companies attacking Tesla and fear-mongering with public posts claiming it is unsafe.

Self-driving is an AI problem, not a sensor suite. “But it has LiDAR, that makes it safer.” As someone who uses Tesla FSD V14 daily, I can say with a high degree of confidence that it would have reacted quicker and not hit the vehicle.

Tesla Optimus vs Figure… it won’t be close. Anyone doubting @Tesla_Optimus at this point is a head-scratcher for me personally. While we have not seen Gen 3 Optimus yet, I truly believe it will be next-level and superior. Here’s why I think Tesla’s humanoid will be better manufactured, more intelligent, and better at tasks. Figure has not “out-collected” Tesla. People who think Tesla hasn’t been collecting data over the last 5 years are, I’m sorry, delusional. Figure runs these “hype videos” because they need funding—it’s really that simple. Not saying what they’re doing isn’t important; I’m just stating the obvious. There are many factors to humanoid robots, and Figure is outmatched by Tesla in every single one. Manufacturing: Figure plans to hit 12,000 a year with their BotQ line and is targeting 100,000 by 2029. Tesla’s Fremont lines are already underway with an annual capacity of 1M, which is 10x Figure in the near term, and the Giga Texas factory is under construction with an annual production capacity of 10M—that’s 100x Figure. Compute: Tesla has Cortex and Cortex 2 with over 500MW of compute. They also work closely with SpaceX, which has the most compute in the world. Figure has about 5% of Tesla’s Cortex compute and wants to invest $1B in future infrastructure. They also lease compute, which is very expensive; Tesla can do this in-house. Frontier model: Tesla will use Grok, now a leading frontier model that could be #1# by year-end. This is one of the most important aspects—without intelligence, your bot is almost useless. Tesla will also use a neural net with massive training data. Figure has its own (Helix), but I don’t see it competing (though that could be up for debate). Supply chain: Do we even need to get into that? Tesla obviously has far more experience than Figure—it’s not even close. Chips: Figure does not design or ship its own inference chipset. They use NVIDIA RTX-class GPU modules in the torso. Tesla will use its own chipset (AI5 and beyond) and will fab its own chips with the Terafab. Not sure how you outcompete that. NVIDIA cannot support millions of robots for just Figure. In conclusion: My prediction is Tesla will mop the floor with companies like Figure, and it won’t be close. There’s plenty of room for other competitors in the humanoid space and maybe they succeed, but Tesla’s scale is in a league of its own.

I know this a very exclusive event but @DavidMoss deserves to go imo. He showed a massive amount of people inside and outside of the Tesla bubble what this FSD is capable of. Multiple coast to coast drives and a 26,000 mile+ streak. His dad is also blind and in a way the Cybercab is a vehicle that will help him get around wherever he needs to go in the near future. If I sound like I'm shilling, well I am. @robotaxi @Tesla

I love not being able to livestream or upload videos correctly on the best platform out there. Not being able to log in on my Mac or other PC is ridiculous. What's wild is yesterday everything was working; today nothing works. Had some cool content planned for today, but I guess f me. What's the point of two-factor authentication if it doesn’t work? And neither dies the back up code. Send email code to sign in? Nope, doesn’t send the code. Send phone number code to log in? Nope, doesn’t work. I’ve tried clearing cache and a multitude of other things, but nothing works. Just says “an unexpected error has occurred, please try again.” We need a live chat we can talk to. Make a Grok version and then escalate to human if needed. Crazy I can’t produce content because I literally can’t LOG IN!!! This also means I can’t go live, which sucks. I’ve also almost finished my live streaming and recording studio where I plan to do live streams in the future. I can’t lie, extremely frustrated @Support @X @XCreators

Gary’s going to be confused by the end of 2026 when there are around 1,000 Robotaxis in the Tesla fleet. He’s going to be extremely confused next year when there are thousands more and likely take the lead for AV’s. Tesla isn’t slowing down robotaxi they’re ramping faster than any other AV operator has ever ramped since its inception. It only seems slow because they’re at the beginning of the exponential growth curve.

Will we have student drivers testing in 5 years? Yeah, most likely. Even with the scale of robotaxis, AI, humanoids, and self-driving vehicles, I still think there will be a lot of manually driven cars. Will babies born this year in 2026 need to learn to drive in 2041? I highly doubt it. There will still be manually driven cars, but also millions of robotaxis and millions of humanoids. For some people it won't make sense to own a car, especially one that doesn’t drive itself. Personal autonomy will be very abundant, and by then the script will have flipped from “I don’t trust self-driving cars” to “I only trust self-driving cars.” Our world is going to change so much in the next 5 years but change exponentially in the next 15.

I will selfishly say I think premium plus subscribers should be able to use Grok Bot but I’m sure it will get released to us eventually. However the application of this is incredible. Being able to assign tasks in the background and even teach it a new skill it can then do automatically. Really awesome to see Grok evolving from just an LLM.

The ability to make a colored light bar will definitely be useful for robotaxi, to help let people know which Cybercab is the right one to get in. They could flash a certain color or make a certain pattern! I know other AV like Waymo and Zoox do this in their own way too.

The cart corral strikes again. V14 in particular has a huge weakness with stationary objects behind the vehicle. Idk if it’s a lapse in training data or what. I actually would assume so because V14 lite also has this issue and it was distilled from V14. Definitely something I hope is a part of one of those 7 new improvement tracks with V15.

There has been some controversy about Tesla limiting their chargers on the website to Tesla owners only. I don’t see an issue with this. They’re likely doing this because supply is in high demand and they want to prioritize their customers first, which they should. It’s hilarious to see people have a problem with this, as if every other company doesn’t prioritize themselves the same way. Imagine a brand-new owner being told by Tesla, “Sorry, we’re sold out of the Wall chargers because we gave those to other EV owners.” That makes no sense. If you want a Tesla charger, buy a Tesla. If not, go to Lowe’s or Best Buy and get one there.

There will be two eras in the near term for Robotaxi: pre-Cybercab and Cybercab. Every other AV operator has their official Robotaxi vehicle for service. Tesla is just using Model Ys, the same ones I use every day to get around. Once Tesla releases the Cybercab into the fleet, everything changes. It is Tesla’s first vehicle built from the ground up as a fully autonomous Robotaxi. Cost per mile goes down, efficiency goes way up, no controls, no HD mapping, with an annual run rate of over 125,000 (based on Q2 deck) that will increase to over 1M over time. It is the safest, most affordable vehicle you can travel in, and it’s not even close. People are underestimating this vehicle.

Game changer for robotaxi. In 2028 when the summer Olympics are in LA Tesla @robotaxi should livestream the Olympics in Cybercabs so you never miss anything. Make an exclusive deal or do whatever you have to do. The experience will be unbeatable, there will be nothing like it

The best part about Tesla FSD is that one day every car they make will be 100% fully autonomous and there is quite literally nothing any hater or doubter can say or do about it. I still believe AI4 will accomplish Unsupervised for personal vehicles, although there will need to be legislation. AI5 and beyond will just be basically “perfect”. Also will be used for inference among other things. The future is going to be amazing.

Sure @Apple I switched to Apple one family plan because I was paying what I thought was way too much for Spotify duo and Apple TV separately. Why are we going up by $2 you guys spending too much CapEx on Siri? 😂