Farzad: Tesla Says Robotaxi Drove 3.1 Million Miles and Injured No One + Tesla Stock "Literally Changed My Life"

Sunday at 3:05am CT, Farzad posted a video titled Tesla Says Robotaxi Drove 3.1 Million Miles and Injured No One. On Saturday, Elon gave GB300 chip counts for four AI facilities, and Farzad answered Elon's post on Tesla and SpaceX wealth: "My Tesla stock literally changed my life!!!"

"Injured No One": Tesla's robotaxi number

Farzad, Sunday at 3:05am CT: Tesla Says Robotaxi Drove 3.1 Million Miles and Injured No One

That's the title of a video he posted. The numbers come from Tesla's new Robotaxi Safety Report on its website. Tesla says its Robotaxis drove 3.1 million miles with no safety monitor aboard from January 22 to October 18, 2026, in Texas and Florida cities, with no one injured, no Robotaxi airbag deployments and no pedestrians or cyclists hit. Tesla only counts an injury when the Robotaxi made first contact. It isn't a zero crash claim: Tesla also lists the crashes it reported to federal safety regulators.

Ethan McKanna, Saturday at 4:04pm CT: The price delta isn't this crazy all the time, but man it's so nice having Robotaxi be faster and cheaper

Ethan added: I haven't used Uber/Waymo in a while now and really never want to again while it's still on the Uber app

The post includes an image that this pull doesn't describe, so the actual prices aren't here. Farzad liked it and RTd it at 4:26pm CT.

Robotaxi Tracker, Saturday at 7:15pm CT: Halloween in Austin: Cybercab beat Waymo every hour from noon to 7pm, hitting a record 25.9 sightings per 1,000 camera snapshots at 4pm

That number counts how often the cars appeared in camera snapshots. It isn't a count of rides or of fleet size, and the post doesn't say whose cameras these are or where they point. It also includes an image this pull doesn't describe. Waymo is the robotaxi service owned by Google's parent, Alphabet. Farzad liked the post and RTd it Sunday at 6:38am CT.

Check Farzad's video here. Ethan McKanna here. Farzad's RT of Ethan here. Robotaxi Tracker here. Farzad's RT of the tracker here.

Neuralink's R1 robot, up close

Cleo Abram, Saturday at 12:44pm CT: Full HUGE* If True episode on Neuralink and the future of brain computer interfaces

She went on: I toured @neuralink and got a close up view of their new R1 surgical robot, to see the future they and @elonmusk are building and how it is already changing lives...

A brain computer interface is a device that reads signals from the brain and turns them into commands for a computer. The full episode is attached to the post, and the post lists its chapters. Some of them:

• How do you get a chip into your brain? (5:19)
• What's inside Neuralink's implant? (8:07)
• How does robot do brain surgery? (11:07)
• How are brain computer interfaces already being used today? (15:05)

Those are chapter titles. The post doesn't say whose lives the implant is changing or how. Farzad liked it.

At 2:37pm CT, Farzad posted his own video, titled Elon's Neuralink Just Made An Insane Breakthrough. This pull only has the title.

Check Cleo Abram here. Farzad's video here.

Colossus 2: Elon's GB300 count

Elon, Saturday at 12:45pm CT: Macroharder and Minihard both have 220k GB300s with 800 gig nics configured for training. Gun Range will have the same. Spam (used to house Hormel products) will have ~200k GB300s, but configured for inference, so much simpler networking and less than half the fiber.

He added: I walked the whole Colossus 2 site around midnight and again this morning. Good progress by the SpaceX team and contractors. Gun Range is tracking to be operational by end of November and Spam in December.

GB300s are Nvidia AI chips. NICs (network interface cards) connect each server to the network, and "800 gig" means 800 gigabits per second. Training is how an AI model gets built: a huge number of chips work on one job and pass data back and forth constantly. Inference is running a finished model to answer requests. Elon says that's why Spam needs much simpler networking. Hormel is the company that makes Spam, the canned meat.

By Elon's numbers, Macroharder and Minihard have 220,000 chips each. Gun Range gets the same by the end of November, and Spam gets about 200,000 in December. That adds up to roughly 860,000 GB300s. The counts and timelines are Elon's.

He was replying to herbertong, 10:35am CT: Macroharder takes up about 810,000 square feet. Minihard is about 312,000, and @CernBasher says it packs in a lot more compute despite the far smaller footprint. herbertong adds that Macrohard has about 550,000 GB200s and Macroharder about 220,000 GB300s, and that a land swap clears space for a fifth data center, with a warehouse across State Line Road that could become a sixth.

Farzad liked Elon's post and RTd it at 3:23pm CT.

Check Elon here. Farzad's RT here. herbertong here.

"The best construction company in the world"

Andrea Esposito, Saturday at 8:16pm CT: Talked with someone in the know about SpaceX / xAI construction roadmap last week. I am now convinced they will single-handedly accelerate 10x the US infrastructure building capabilities, through aggressive vertical integration + applying Elon's algorithm.

The post goes on to list what Andrea expects them to build:

• data centers
• rockets and satellites factories
• launch towers
• LNG/LOX production plants and pipelines
• space data centers factories
• eventually, lunar and martian infrastructure

Then: Construction will become one of their biggest competitive advantages, and they'll become the best construction company in the world. They'll also deploy compute faster and more efficiently than anyone else, and rent out inference to the model makers.

And on AI models: I don't know if this means they will also win on models, but Elon will definitely try. Him who controls the weights controls the culture.

All of that is Andrea's opinion. The post doesn't name the source or say what's on the roadmap. Vertical integration means making more of what you need in-house instead of buying it. Elon's algorithm is his five-step engineering process: question every requirement, delete parts and steps, simplify, speed up, then automate. LNG is liquefied natural gas and LOX is liquid oxygen. Weights are the learned numbers inside an AI model.

Andrea was quoting Object_Zero_ from 2:59pm CT. That post put the global economy at $128 trillion and global capital expenditure at $28.5 trillion, with $3.9 trillion in manufacturing and $3.6 trillion in residential property. The post lists data centers and compute at $950 billion and argues that money isn't the problem for data centers. Getting them built on time is. Capital expenditure (capex) is money spent on long-lived things like buildings, machines and chips.

Sunday at 5:40am CT, Elon replied to an October 26 post from Milk Road Stocks: Our capex efficiency is greatly underestimated

The Milk Road post quoted Kyle Reidhead: "SpaceX is on track to spend $470B on capex in the next 5 quarters, ~2x what Amazon, Google or Meta each spend." It says he argues SpaceX can fund that because it builds so fast: "Elon can do this in sixty days, ninety days," and "within six months, it's already paid back." The $470B figure comes from Reidhead, and Elon's reply neither confirms nor disputes it. Farzad liked Elon's reply and RTd it at 6:21am CT.

At 6:34am CT, Elon replied to Andrea: Yup

Farzad liked Andrea's post and Elon's reply, and RTd Andrea at 6:39am CT.

Check Andrea Esposito here. Farzad's RT here. Object_Zero_ here. Elon's capex reply here. Farzad's RT here. Milk Road Stocks here. Elon's reply to Andrea here.

Tesla and SpaceX "created meaningful wealth"

Elon, Saturday at 10:24am CT: Tesla and SpaceX have created meaningful wealth for hundreds of thousands of employees over time and improved the retirement savings of hundreds of millions of people

That's Elon's claim. The post doesn't explain how he arrives at either number. It quotes an October 1 post from TeslaLarry that this pull shows only as a link.

Farzad liked it and RTd it at 11:33am CT. A minute later he replied: My Tesla stock literally changed my life!!!

He's describing his own experience, and he doesn't put numbers on it here.

Check Elon here. Farzad's RT here. Farzad's reply here.

Start with dr eggbot

Ed Tadros, Saturday at 1:17pm CT: I think every Grok Bot install should start with dr eggbot ( a nod to Dr. Eggman?). I had a few friends just get started on Grok Bot, and this is the first bot I tell them to use.

Ed was quoting poteto, 12:01pm CT: here's a short grok bot guide on how i create high quality bots and manage my team.

poteto calls dr eggbot "my bot designer that i use to create bots" and adds "you don't need a lot of bots." poteto says a few good bots can be more useful and easier to manage than a large team. Farzad liked Ed's post.

Check Ed Tadros here. poteto here.

"You're likely not pushing it enough"

Aaron Levie, Saturday at 5:02pm CT: This is one of the better litmus tests for whether you're pushing your use of AI enough or not.

Levie writes that up until "maybe six months ago," the best you could do with AI was make what you already would, "just doing it a little faster." Now, Levie says, working with AI "can and should dramatically exceed" what you could do before, "in a fraction of the time." The post closes with the test: if you could have done what you're doing before AI, you're likely not pushing it enough.

Levie also writes: We're unquestionably entering a period where AI will be smarter than us on most axes. The post adds that this won't hold on every dimension, because "intelligence is still very jagged." In other words, AI is strong at some tasks and weak at others.

Levie was quoting sundeep, 11:25am CT: If you're going to make something with AI, make sure it's 10,000x better than anything you would have made yourself. If not, you just come across as lazy not smart :)

Farzad liked Levie's post and RTd it Sunday at 6:28am CT.

Beff (e/acc), Saturday at 6:36pm CT: Jevon's paradox of content production. Democratizing access to AI tools and lowering friction for artistic workflows increases total amount of quality content produced by artists.

The Jevons paradox is the old observation that when something gets more efficient to use, people end up using more of it in total. Beff was quoting ashebytes, who wrote "now you only have to be good at talking to the camera" while watching their agents make "all these cute edits, faster and faster." Farzad liked Beff's post.

Dave Lee, Saturday at 10:36am CT: I remember ages ago when OpenClaw was all the rage.

Farzad liked it and replied at 3:42pm CT: The start of the agentic revolution.

Agentic AI means AI agents that carry out tasks on their own. Neither post says what OpenClaw is or when it was popular.

Check Aaron Levie here. Farzad's RT here. sundeep here. Beff here. Dave Lee here. Farzad's reply here.

Farzad's videos

Farzad posted five videos between Saturday at 9:11am CT and Sunday at 7:10am CT. This pull only has their titles. He replied to each one with the same link to Shorty, a Grok Bot of his whose link card says it "Owns YouTube Shorts end-to-end." Three replies name Shorty. The replies under the Neuralink and Starlink videos say "Made with my Grok Bot Solo" instead, but the link still opens the Shorty card.

Under the Neuralink video at 6:54pm CT Saturday, Luke James Stephens (borisstephens) suggested Farzad "upload a different ending for 𝕏 that doesn't plug the YouTube algorithm as it's the wrong context." Farzad liked it and replied at 10:57pm CT: Great call. He doesn't say whether he'll change the endings.

Check Why a Pig Kidney Kept a Man Off Dialysis for 271 Days, Saturday at 9:11am CT, here.

Check Elon's Neuralink Just Made An Insane Breakthrough, Saturday at 2:37pm CT, here.

Check Why SpaceX Is Retiring the Rocket That Just Flew 700 Times, Saturday at 9:02pm CT, here.

Check Tesla Says Robotaxi Drove 3.1 Million Miles and Injured No One, Sunday at 3:05am CT, here.

Check How Starlink Actually Works, Sunday at 7:10am CT, here.

Check borisstephens here. Farzad's reply here.