Farzad: Don't Sleep on Zuck & Meta + The X Subscription + Sacks on Anthropic's Distillation Defenses + Robotaxis vs Speeding

Farzad posted a long original Monday night arguing it's Anthropic + SpaceX + Meta long term, with OpenAI playing catch-up. Farzad also liked posts on the new X Subscription, RTd David Sacks on Anthropic's distillation defenses and Elon on SpaceX vs China, and flagged a Grok Bot outage.

"Don't sleep on Zuck & Meta"

Farzad, Oct 12 late evening CT: Don't sleep on Zuck & Meta. Metric-shit-ton of compute, clearly super talented team, and a CEO that has learned a lot of lessons and doesn't give a shit anymore.

Farzad's long-term call: I think it's Anthropic + SpaceX + Meta long term. Farzad added Google as a maybe for later.

Then Farzad asked where OpenAI fits in this whole equation. Farzad's answer: OpenAI can't beat out Meta's distribution. They can't beat out Anthropic's models. They can't beat out SpaceX's compute build-out. They can't beat Google's data moat.

Farzad raised physical + digital as one way OpenAI could differentiate, then named who it would run into there. One is Apple, if it wakes tf up from its giant slumber and decides to actually make something interesting for once. The other is Tesla + SpaceX with Optimus + Cars + whatever else they build long term in the physical world.

Farzad also wrote that with their march towards the frontier, OpenAI seems to be getting into some serious PR disasters with Huggingface etc. Farzad's close: Just feels like they're entering a period where they are playing catch-up in everything.

That's Farzad's projection and Farzad's read on OpenAI, not reported news.

Check Don't sleep on Zuck & Meta here.

The X Subscription

Farzad liked SpaceXSI, Oct 12 evening CT: We've unified plans across Grok, Cursor, and X Premium. Introducing The X Subscription.

Farzad also liked SawyerMerritt, Oct 12 evening CT: SpaceX has just officially introduced a new "X subscription" that combines Grok, X and Cursor into one plan. Sawyer listed four tiers: Plus at $8/month, Premium at $30/month, Super at $100/month and Ultra at $200/month.

According to Sawyer, Plus includes Cursor access, Grok Chat, Voice, Imagine & Build, the Grok API, a shared usage pool across products, the Premium checkmark, fewer X timeline ads and creator monetization, plus other X features. Premium adds 4x more usage compared to Plus, Grok Bot, the latest Grok models, faster Chat responses, AI images & videos, ad-free scrolling on most surfaces on X, a preferred @handle and a higher reply boost. Super gets 15x more usage compared to Plus and priority model access at peak times. Ultra gets 50x more usage compared to Plus, earliest access to all products & models and the fastest Chat responses & generations.

On Sawyer's list, Grok Bot starts at the $30 Premium tier. The tiers and features above are as Sawyer listed them.

Farzad also liked poteto, Oct 12 evening CT, quoting the SpaceXSI post: it's finally here! one unified sub with incredible value across Grok, Grok Bot, Cursor, and X Premium

Check SpaceXSI here. SawyerMerritt tiers here. poteto here.

Sacks on Anthropic and distillation

Farzad RTd and liked DavidSacks, Oct 12 late evening CT. Sacks wrote that Anthropic says it has reduced the data harvested from Claude in distillation attacks by around 90 to 99%, and congratulated Anthropic on the achievement.

Sacks tied it to a point from All-In Ep. 282 a few months ago: Anthropic, not the government, was in the best position to stop distillation of its own models. Once the horse is out of the barn, there's limited value in after-the-fact government action.

Sacks recalled that at the time there was a major push in Washington to restrict open models on IP grounds. Per Sacks, 50 companies, including Nvidia, Microsoft, Meta, IBM, Palantir and Dell, published an open letter in defense of open-weight models. Jensen Huang promoted the letter in a first-ever X post, and nearly 200 startups sent a similar letter.

Sacks wrote that restricting open models risked removing the incentive for closed-model companies to solve the problem themselves. In Sacks' view, restrictions could also reward negligence by stifling competition while those companies kept collecting revenue from distillation. The administration did not impose those restrictions. Sacks' close: Now we see the result. Anthropic invested in defenses and they worked. Not every technical challenge is a reason to run to Washington for new regulation.

The 90 to 99% figure is Anthropic's, as Sacks relayed it. The policy argument is Sacks'.

Check DavidSacks RT here. DavidSacks here.

SpaceX as "an anomaly in the matrix"

Farzad RTd and liked Elon, Oct 12 early afternoon CT. Elon, Oct 12 afternoon CT, was replying to MelvinInvests, Oct 12 late morning CT: This selloff in memory stocks makes absolutely NO sense to me. $MU, $SNDK, Samsung and SK Hynix are getting hit after China's CXMT announced progress on its new 4F² DRAM architecture, with plans to introduce DDR5 server memory using the technology by the end of 2026.

Elon: When China figures out how to manufacture a product, their ability to scale is often greater than the rest of the world combined. We saw this with lithium batteries and solar, for example. SpaceX is a rare counter-example, where its mass to orbit is roughly an order of magnitude greater than the rest of the world, BUT China now does a majority of non-SpaceX mass to orbit. SpaceX is an anomaly in the matrix.

On memory, Elon wrote: Maybe RAM will be different, but RAM was a low margin commodity only ~3 years ago. LP5 and LP6 are designed to a common standard that China seems to be solving fast. Elon expects some things to be harder to solve: newer HBM architectures, an SRAM layer below logic, and 3D stacking RAM on top of logic while keeping the logic cooled.

The mass to orbit comparison is Elon's. The CXMT plans are as MelvinInvests summarized them.

Check Elon RT here. Elon here. MelvinInvests here.

"Robotaxis will solve speeding," plus Grok Bot on FSD

Farzad RTd and liked ALScyborg, Oct 12 evening CT. ALScyborg, Oct 12 afternoon CT: Robotaxis will solve speeding. I have ALS and I can't drive anymore, so now I ride along and watch other people drive.

ALScyborg did the math. Driving 10 miles at 70 mph instead of 65 saves about 40 seconds. In town, 5 miles at 45 instead of 40 saves less than a minute, and a single red light can wipe that out.

ALScyborg's theory is that we speed because we are bored and want the drive to be over. ALScyborg pointed to research that links boredom proneness to risky driving. According to NHTSA, as cited by ALScyborg, speeding was a factor in 29% of US traffic deaths in 2024, or 11,288 people. ALScyborg: AI doesn't get bored. It knows 5 mph faster doesn't matter. ALScyborg added that Waymo says its cars follow posted speed limits except in extraordinary circumstances. The close: When nobody is bored behind the wheel, nobody needs to rush. The future is calm, safe and on time.

Farzad liked ArthurMacwaters, Oct 12 late evening CT: connected grok @bot to my tesla while using FSD. incredible. we are living in the future

Check ALScyborg RT here. ALScyborg here. ArthurMacwaters here.

Grok Bot messages not going through

Farzad, Oct 12 evening CT: Is Grok @bot acting weird for anyone else? Messages not going through.

About nine minutes later, Farzad posted: Grok Bot experiencing a major outage ATM.

Later that evening, Farzad liked Elon's reply to Kevin_Logan: Fix coming within a few days. Kevin_Logan, Oct 12 late afternoon CT, had posted: Thank you again everyone who bookmarked and shared this. I hope it helped your Grok @bot team as much as it has mine! Farzad replied to Elon: YES! Elon's reply doesn't say what the fix is for, so it isn't clear it's tied to the outage.

Check Grok Bot acting weird here. Grok Bot outage here. Elon here. Kevin_Logan here. Farzad's reply here.

New Shorts

Farzad posted two new Shorts, one Oct 12 evening CT and one Oct 13 morning CT. Each one is followed by a "Full video" reply that links the long-form YouTube video.

Check Starship Is Really a Freight Elevator to Space here. SpaceX Almost Died in 2008. Look at It Now here.