Farzad: "$SPCX Core Competency Is Working Around Massive Bottlenecks" + "AI Satellites in Space About to Accelerate 1000x"

Wednesday at 7:43am CT, Sawyer Merritt posted that SpaceX officially announced a partnership with Qualcomm to help build Starlink Mobile's nationwide terrestrial 5G network. Farzad liked and RTd it. On Tuesday he'd written that "$SPCX core competency is working around massive bottlenecks," and that evening, at 8:46pm CT, he posted: "AI satellites in space about to accelerate 1000x."

SpaceX picks Qualcomm for Starlink Mobile

Sawyer Merritt, Wednesday at 7:43am CT: Big news! SpaceX has just officially announced that they are partnering with Qualcomm to help build @Starlink Mobile's nationwide terrestrial 5G network, with deployment beginning in 2027.

He goes on: "Under the multi-year agreement, SpaceX plans to deploy tens of thousands of cellular sites across the U.S., powered by Qualcomm's Dragonwing infrastructure technology." He lists the QRU100 and QDU100 platforms and 64T64R Massive MIMO, and says SpaceX's goal is "the first fully integrated nationwide satellite and terrestrial mobile network." He calls it a step toward SpaceX competing with AT&T, Verizon and T-Mobile, which is his read. Qualcomm's own release says the same: tens of thousands of cell sites, rollout starting in 2027, and "our biggest 5G deal yet." Neither company gave a dollar figure. Terrestrial means ground-based, so this is cell sites on land and not satellites. Farzad liked the post and RTd it at 8:12am CT.

Yesterday's Digest covered The Washington Post's report that the possibility of a SpaceX mobile network had rattled major carriers, and this post goes further, naming a partner and a start year.

The day before, Tuesday at 1:08pm CT, Farzad quoted Elon: $SPCX core competency is working around massive bottlenecks.

Elon's post, 10:07am CT: And electrical equipment, cooling, networking and batteries

Elon was quoting euwyn, 5:00am CT: Elon Musk says the AI boom will hit a WALL by the end of 2026.

euwyn continues: "He says the whole industry is about to learn a hard lesson in hardware. Chip output is exploding, but electricity is nearly flat everywhere outside China. So the two lines cross this year." It goes on: "He says by the end of the year, we will be making more AI chips than we can actually turn on." It ends: "In the end, the AI race comes down to one thing: Raw electricity." euwyn credits the remarks to Elon on Dwarkesh Patel's podcast. That's euwyn's summary, not Elon's exact words. Elon's reply reads like he's adding to the list of things in short supply. Farzad liked Elon's post.

SPCX is SpaceX's stock ticker. The shares have traded on Nasdaq since June 12. Farzad's line is his opinion. He doesn't say which bottlenecks he has in mind or how SpaceX works around them.

Blake Scholl, founder and CEO of Boom Supersonic, quoted the same euwyn post at 2:53pm CT: TL/DR need more turbine

Farzad liked it and RTd it at 2:55pm CT. Eleven minutes later he posted a video titled Elon Says Electricity Outside China Is Basically Flat (listed below).

Check Sawyer Merritt here. Farzad's RT here. Farzad's $SPCX post here. Elon here. euwyn here. bscholl here. Farzad's RT here.

"Simply unstoppable"

Farzad, Tuesday at 10:06pm CT, called it an "interesting thought experiment." His question: if a change in who controls Congress brings in new lawmakers who want to slow AI down to keep voters happy, how far could that go? He puts the range anywhere from "a little slowdown in build outs" to "full on war."

He calls a full-on war "highly unlikely" and gives three reasons. It "would literally destroy the economy," it's "the dumbest national security move of all time," and "too many people use it now and find it genuinely great."

His read: "even though it's gonna get choppy and weird, the country's AI progress is simply unstoppable." He argues AI is "WAY too embedded in defense, the FBI, CIA, and other high-caliber orgs," and says there's "zero shot those groups would ever be on board with slowing it down considerably."

His guess is that "one or two projects get shut down because of politics," but "growth continues to accelerate." He names one effect "for sure": "the acceleration of AI satellites in space, which completely bypasses all this nonsense."

He also brings in China, writing that "the US cannot afford to actually, meaningfully, slow down AI in any way, shape, or form," or else "China will gain techno-supremacy." His summary: AI is too big to fail.

On jobs, he calls AI "a mega-gigantic productivity accelerant" that "could lead to job loss." He expects that "in pretty significant ways starting next year." He adds that it also creates "a massive amount of new opportunities for people to embrace the technology."

All of that is Farzad's projection. He doesn't name the projects he expects to get shut down, and apart from "next year" on jobs he gives no timeline.

EntropyCowboy replied at 10:09pm CT, agreeing it "might be too late" for a total war on AI but saying robots are another story: "for robots, man, we are fucked."

Farzad, 10:13pm CT: Includes robots IMO. I don't think physical/digital AI slow down at all.

Check Farzad's thought experiment here. EntropyCowboy here. Farzad's reply here.

"Accelerate 1000x"

Farzad, Tuesday at 8:46pm CT: AI satellites in space about to accelerate 1000x.

That's his projection. He doesn't say what the 1000x measures or how long it takes. At pull time it had more likes than anything else he posted in this window.

Farzad also liked two other posts making the same point. Dylan Patel of SemiAnalysis, 8:01pm CT: I am now bullish space datacenters

It includes an image this pull doesn't describe. Elon, Wednesday at 12:12am CT: Data centers in space looking pretty appealing right now

On data centers on the ground, MF_Camillus posted Tuesday at 11:02pm CT: The idea that the US midterms will somehow derail the AI data center buildout strikes me as another deeply flawed narrative

He gives his reasons in a long post. Farzad liked it and quoted it Wednesday at 8:23am CT: Really good.

Scholl, Wednesday at 7:48am CT: Winning data center platform: 1/ no subsidies, no tax breaks 2/ make your own electricity, waterlessly 3/ equipment should be essentially inaudible at the fenceline

The fenceline is the edge of the site, where neighbors would hear any noise. That's Scholl's checklist, and the post doesn't say whether any current project meets it. Farzad liked it and RTd it at 8:09am CT.

Check Farzad here. Dylan Patel here. Elon here. MF_Camillus here. Farzad's quote post here. bscholl here. Farzad's RT here.

"It's ACCELERATING"

Ivan Burazin, Tuesday at 11:05am CT: We just crossed $100M in run rate.

Run rate is current revenue projected out over a full year. He writes: "Four quarters ago, we were at $1M. Then this year alone, we grew 37.5x." Not long before that, he says, the company did a hard pivot: it handed its old customers to competitors, "rebuilt from scratch, and relaunched with a few months of runway left." Runway is how long the cash lasts.

He describes the bet as "models think on GPUs but work happens on CPUs." So, he writes, "every agent that writes code, opens a browser, or runs an RL job needs a composable computer that's isolated, stateful, and ready in milliseconds." RL is reinforcement learning, a way of training AI by trial and reward. In his words, going from $1M to $100M in about a year "puts us in the same cohort as the likes of Anthropic and Cursor."

The numbers are his. The post doesn't name the company, but Burazin is CEO and co-founder of Daytona, which rents out cloud computers for AI agents. Daytona's blog posted the same figures that day. The post includes a video this pull doesn't describe.

Farzad liked it and quoted it at 2:50pm CT: The economy is transforming right before our eyes.

And: And 99% of people have no idea that it's ACCELERATING!!!!!

The 99% is a figure of speech, not a survey.

Check Ivan Burazin here. Farzad's quote post here.

Shawty: about 18 cents a video

Farzad, Tuesday at 11:15am CT, quoting his request post from Monday at 6:25pm CT: This Grok Bot experiment went WAY better than I expected!

His stats: "Each video cost ~$0.18 to generate. Average time from prompt to uploaded video was about 30 minutes." The original post promised a reply "within 20 minutes," so by his own numbers the average came in about ten minutes over.

Here's how it worked, per Farzad. Shawty "ran the entire thing end to end." It had a session of Opus 5.5 medium, "which did all the creative work and animations," and a session of GPT 5 astra medium, "which wrote the scripts." Grok Bot "did the rest, from Orchestration, to research, uploading to X, and everything in between."

He says Shawty "is optimized for speed," and that you can "go maximum quality by upping the thinking efforts and/or asking for much more complex animations."

Space_Me_Now asked whether the 18 cents covered "just grok part" or "all the farmed out stuff too." Farzad, 2:18pm CT: .18 for the whole thing.

steveMmattison asked: if a video "only cost you 18 cents in tokens, why TF am I running out of GB tokens for building websites & apps & sending emails?" Farzad, 2:17pm CT: I spent A LOT of time optimizing.

The cost is Farzad's figure. He doesn't break it down by model or explain how he added it up.

Check Farzad's stats post here. The original request post here. Space_Me_Now here. Farzad's reply here. steveMmattison here. Farzad's reply here.

"Grok Bot."

benln, Tuesday at 8:01am CT: Best AI product of 2026?

Farzad quoted it at 10:29am CT: Grok Bot.

Peter Diamandis, 9:28am CT: Got to say it. GrokBot is AWESOME. Actually addicting.

Elon quoted him at 10:11am CT: Grok @Bot is like hiring a super competent employee. Treat Bot like you would a colleague.

Farzad liked both posts. At 10:26am CT he replied to Elon with "It's insane:" and a link to his Shawty request post.

Sawyer Merritt, 1:17pm CT: Grok @Bot's new calling feature is very impressive.

He explains: "You just tell it who to call and why, and it handles the conversation for you." It "shows a real-time transcript of the call," and if Grok needs more from you, "a little answer box pops up where you type/select your response, and Grok relays it to the person on the phone." He calls it "Pretty incredible" and attaches an example with the call audio, which this pull doesn't describe. That description is Sawyer's. Farzad liked the post.

The Bot account, 3:50pm CT: Try the new shopping experience in Grok Bot.

It adds: "Find the best deal, compare items, and have Bot track all your orders." Elon quoted it Wednesday at 2:16am CT: Grok @Bot shopping is great

Farzad liked Elon's post.

Check benln here. Farzad here. Peter Diamandis here. Elon here. Farzad's reply here. Sawyer Merritt here. The Bot account here. Elon on shopping here.

"Internet is going full avatar"

Farzad, Tuesday at 7:35pm CT: OK I think Gemini 4 Argon on high might be the best writer.

That's his opinion. He doesn't say what he compared it against or what kind of writing he means. His Shawty setup, described the same day, used GPT 5 astra for scripts.

Steven Mark Ryan, 6:15pm CT: There will be an INSANE proliferation of EXTREMELY excellent content produced ~entirely by Super Intelligence.

He calls this "inevitable," but adds that "People will put an increasingly higher value on 'real' human content as a boutique novelty." Farzad liked it and replied at 7:36pm CT: I think humans will get "real" human content outside instead of a screen.

And: By then, it'll literally be IMPOSSIBLE to tell human from AI content. Internet is going full avatar.

Ryan, 7:39pm CT, agreed that "IRL is gonna have a big comeback" but said people "will also put a big premium on the 0.01% of pure human content that remains." Farzad, a minute later: How will you know it's pure human?

Check Farzad on Gemini here. Steven Mark Ryan here. Farzad's reply here. Ryan's reply here. Farzad's question here.

Roadster POV

Sawyer Merritt, Tuesday at 10:51am CT: Insane POV of the next-gen Tesla Roadster doing a kickflip, going up the ramp and 0-60 mph test.

He adds: "That shot of the rear rocket thrusters in the first clip looks awesome." Farzad liked it and RTd it at 10:55am CT. The descriptions are Sawyer's. This pull doesn't show the video, and the post lists no times or specs.

Check Sawyer Merritt here. Farzad's RT here.

Farzad's videos

Farzad posted eight videos between Tuesday at 12:03pm CT and Wednesday at 7:22am CT. This pull has only their titles, so the numbers below come from the titles and the math isn't shown. Under seven of them he replied "Build your own short form content with my Grok Bot Shorty," linking to Shorty's card. Shorty is a different bot from Shawty. Under the Colossus 2 video he replied "Made with my Grok Bot Solo."

Check Elon Says a Space Radiator Sheds 1,300 Watts per Square Meter, Tuesday at 12:03pm CT, here.

Check Elon Says Electricity Outside China Is Basically Flat, Tuesday at 3:06pm CT, here.

Check Why SpaceX Fishes Its $6 Million Nose Cone Out of the Ocean, Tuesday at 6:11pm CT, here.

Check Elon Says Optimus Will Cost Less Than a Car. Here's the Math, Tuesday at 9:01pm CT, here.

Check Why the Starlink Dish Aims at Satellites Without Moving, Wednesday at 12:10am CT, here.

Check Elon Says Starship Will Carry 100 People to Mars. Here's the Math, Wednesday at 3:02am CT, here.

Check Why Landing on Mars Is Harder Than Landing on the Moon, Wednesday at 6:05am CT, here.

Check Inside Colossus 2: How Elon Is Wiring 1.2 Million GPUs, Wednesday at 7:22am CT, here.

Check the Solo reply under the Colossus 2 video here.