Farzad: Grok Bot and Grok 5 Will Melt Faces + AI Companies as GDP Accelerators + Cybercab vs Waymo in Austin + Karpathy's LLM Funnel

Farzad posted two long originals Sunday night. One guessed that Grok Bot plus Grok 5 is going to melt people's faces off, and the other argued that AI companies need to be GDP accelerators. Farzad also RTd Cybercab vs Waymo sightings in Austin, Karpathy's LLM capability funnel and nikitabier on moats.

"Grok Bot + Grok 5 is going to melt people's faces off"

Farzad, Oct 11 late evening CT: My guess is Grok Bot + Grok 5 is going to melt people's faces off. If @SpaceXSI is actually successful in training a high-quality engineer with Grok 5 (due to training on in-house engineering data), then pairing it with Grok Bot will enable the creation of thousands of new $1m+ businesses basically overnight. Grok Bot + Grok 5 means basically infinite high-quality digital labor to build really high quality stuff that works incredibly well.

Farzad wrote that it wouldn't be a surprise if it begins materially impacting GDP within 12-18 months: I mean how could it not, right? We can all already see how insanely powerful these agents are. Imagine the models that are specifically trained and optimized to fully take advantage of those harnesses.

That's Farzad's projection. It rests on the "if": Grok 5 has to become a high-quality engineer by training on in-house engineering data. Grok 5 has not been released yet; SpaceXSI has said only that it is in training.

Farzad liked AlexFinn's reply, Oct 11 late evening CT: Every day I'm blown away more by grok bot

Check Grok Bot + Grok 5 here. AlexFinn here.

AI companies as GDP accelerators

A few minutes earlier, Oct 11 late evening CT, Farzad posted: The surest way for AI/SI to be a net-positive for humanity is to ensure AI companies are GDP accelerators.

Farzad wrote that in 12-18 months "we're likely going to have literally Superintelligence," meaning AIs significantly smarter than a human at almost everything digital. Those AIs will be unbelievably helpful but also unbelievably distracting, because AIs "will know how to dispense dopamine better than anything else out there (short of drugs)."

If AI companies prioritize dispensing dopamine, Farzad wrote, the majority of AI earnings will accumulate to the few companies that provide that service. Meanwhile people keep paying with money or attention instead of being productive. Farzad's verdict on that path: Bad, bad place to be.

The other path is for AI companies to deploy earnings in ways that are significantly more productive for society. Farzad's examples were next-gen education, solutions to fatal disease, extremely affordable and abundant HIGH QUALITY housing, and tools/services that give people superpowers (cracked-out agents). Then the world will get REALLY GREAT, REALLY FAST. Farzad's close: if AI companies don't do their part to ensure society wins with them, they will go down with the ship. Don't be greedy.

The Superintelligence timeline and both outcomes are Farzad's projections.

Check AI companies as GDP accelerators here.

Grok Bot in the wild

Farzad liked Elon, Oct 11 late afternoon CT: Seriously, this actually works! Think of Grok @Bot like hiring a super smart employee who works 24/7 and is extremely resourceful. You can use it through your phone effortlessly within minutes. Elon linked the Grok Bot app on the App Store and quoted liam_fallen, Oct 11 afternoon CT: Grok Bot bought me an iPhone 18 Pro Max. It was sold out everywhere. Even Apple said I'd have to wait until the end of October. So I gave Grok Bot a virtual card and told it to find me one. According to liam_fallen, Grok Bot emailed stores, found one in stock, and verified and ordered it.

Farzad liked robotaxi, Oct 11 evening CT: Talk directly to your Grok @bot in Robotaxi. robotaxi was quoting BLKMDL3, Oct 11 evening CT: Grok @bot is now available in @robotaxi! I had my bot order me my usual drink while in a Cybercab.

Farzad liked a reply from thejefflutz, Oct 11 late evening CT, to SawyerMerritt. Sawyer, a few minutes earlier: You can now use Grok @Bot in a Tesla Cybercab. You just sign into Grok once, and your Bots will be available on all future rides, and automatically logged out when you exit the vehicle to maintain privacy. I tried it tonight by adding laundry detergent to my Amazon cart. thejefflutz: Tesla / spacexsi are making it so your digital being is with you at all times... imagine grok bot with all of its capabilities inside of your own Optimus in the future

Check Elon here. liam_fallen here. robotaxi here. BLKMDL3 here. SawyerMerritt Grok Bot in Cybercab here. thejefflutz here.

Cybercab vs Waymo in Austin

Farzad RTd and liked SawyerMerritt, Oct 11 late afternoon CT. Sawyer, Oct 11 late afternoon CT: Not even joking, it feels like there are more Cybercabs giving rides in Austin right now than Waymos. All these clips were taken within minutes of each other today. That's Sawyer's read from the clips.

Farzad liked TSLAFanMtl, Oct 11 late afternoon CT, quoting that post: The road network in Austin will turn into a tide of gold...people will wonder wtf is going on. We must closely monitor the price of the service in Austin as the volume scales up. This will give us a clue as to their strategy.

Farzad RTd and liked GrumpyTesla_, Oct 11 evening CT. GrumpyTesla_, Oct 11 afternoon CT: There also seemed to be alot of @farzyness edition Cybercabs driving around! Especially on the highway!

Farzad RTd and liked ethanmckanna, Oct 12 morning CT. ethanmckanna, Oct 11 late evening CT: Robotaxis per 1,000 Austin camera snapshots, hour by hour across ACL weekend 2. Cybercab beat Waymo from 2 to 5pm both Saturday and Sunday. That post quoted an earlier ethanmckanna chart, Oct 11 afternoon CT: Per hour comparison of Waymo v Cybercab over the last 48hrs. 4 hour period where Cybercab was ahead, otherwise Waymo iPace remains the most seen AV on the road. Those counts are ethanmckanna's, from Austin camera snapshots.

Check SawyerMerritt RT here. SawyerMerritt here. TSLAFanMtl here. GrumpyTesla_ RT here. GrumpyTesla_ here. ethanmckanna RT here. ethanmckanna ACL weekend 2 here. ethanmckanna 48hrs here.

Karpathy's capability funnel, plus moats

Farzad RTd and liked karpathy, Oct 12 morning CT. Karpathy, Oct 11 afternoon CT, brought back an April post and wrote that the gap in shared understanding of LLM capability is widening: It's now less "two groups of people speaking past each other" and more a sharp funnel.

Karpathy's napkin math: around 6B people (~75% of the population) have barely come in contact with LLMs at all. Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products who see derpy chatbots. Somewhere around 20M people (0.2%) use frontier-grade LLMs professionally in math&code and see first-hand that large, complex projects that used to take weeks/months can now be completed by agents with a prompt. And ~5,000 people (~0.00006%) have access to frontier-grade systems internally.

Karpathy named three drivers. Impact scales with ambition, problem size and horizon. The system is jagged, with capability peaking in domains that are digital, verifiable and economically valuable. Access differs too: free-tier, paid-tier, internal. Karpathy's close: The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time. The tiers and percentages are Karpathy's estimates.

Farzad RTd and liked nikitabier, Oct 12 morning CT. nikitabier, Oct 11 late evening CT: Technical moats are mostly over at this point. If you're building something that is meant to be durable, the only moats now are: proprietary data, network effects, a physical component.

On a lighter note, Farzad RTd and liked vladtenev, Oct 11 late evening CT: Hearing more and more AI researchers are taking up smoking ... superintelligence is near

Check karpathy RT here. karpathy here. nikitabier RT here. nikitabier here. vladtenev RT here. vladtenev here.

Morgan Stanley on Starship Flight 15

Farzad liked SawyerMerritt, Oct 11 evening CT. Sawyer shared a new SpaceX note on Starship Flight 15 from Morgan Stanley's Adam Jonas. Jonas: "Flight 15 (~1 month from now) is shaping up as SpaceX's biggest test towards Starship reusability since the IPO, with Flight 16 (targeted before year-end) potentially an even bigger one."

Jonas wrote that developments over the coming weeks and months could help profoundly evolve investor expectations about SpaceX's role in converting energy into intelligence at scale through physical agents. The note says each Starship launch of Starlink is equivalent to at least 1 month of Falcon 9 launches. Flight 14 deployed 26 Starlink V3 satellites, and Jonas expects a repeat or a potentially larger load on the next flight (Starship fits up to 60). Each Starlink V3 has 10x the downlink capacity of the v2 mini sats that Falcon 9 launches every ~3 days. By the note's math, Flight 15 would equate to 1-2+ months of F9 launches. Sawyer added that Jonas maintained a $300 $SPCX price target. Those numbers and expectations come from Morgan Stanley.

Check SawyerMerritt Starship note here.

New Shorts

Farzad posted seven Shorts, Oct 12 morning CT. Each one is followed by a "Full video" reply that links the long-form YouTube video.

Check Grok Bot Just Edited My Entire Video here. Every AI Expert Agrees the Shift Is Here here. Tesla, SpaceX, and xAI Are Different Limbs of One Body here. xAI Is Already Building the Brain for Optimus here. Vera Rubin: 30x More Agent Throughput here. Why SpaceX Wants AI Data Centers in Orbit here. One Entity That Builds Chips, Robots, Cars and Rockets here.