SpaceX was down 35% since IPO and the unlock ladder is landing exactly as Rik called it on episode 15. Rik and Ben grade 25 episodes of AI calls: Anthropic vs OpenAI, Claude vs Codex vs Cursor, and why DeepSeek's pricing doesn't even show up on the chart.

It's episode 25 — a quarter of a hundred. Just Rik and Ben this week (Luca's off), so they used the milestone to pull 25 lessons out of the archive and grade every one against what actually happened.

IN THIS EPISODE:

• Grading the SpaceX bear case from episode 15 — down 35% from IPO, the $101B unlock ladder landing exactly as called

• Rik cancels the $100 Claude Max plan again, and explains why Cursor plus Codex replaced it

• A developer turned $800 in AI subscriptions into $50,000 of coding value

• DeepSeek's pricing is so cheap it doesn't register on the chart, then a price hike lands mid-episode

• Gavin Baker's supply and demand math: token demand is 10xing, supply is only 3xing next year

• 25 episodes, 25 graded predictions across three storylines: OpenAI vs Anthropic, open vs closed source, and the Elon Corner

• Why Ben thinks we still haven't hit AGI, a year after Manus made everyone think we had

CHAPTERS:

00:00 Cold open, the SpaceX call

00:31 Hello everyone, 25 episodes and three wars

01:26 Ben's Brooklyn life, Rik turns 30 tomorrow

01:58 Claude is now Ben's full-time editor

04:43 The U-turn, content versus maxing tokens

06:46 Alex Finn's ChatGPT Voice hype, graded

10:22 The goals myth, then $50K for $800 in subscriptions

13:16 Opus burns 100x more tokens than Terra

15:02 DeepSeek, too cheap to see on the chart

17:40 Gavin Baker, 10x demand, 3x supply

20:48 The average intelligence thesis, restated

22:32 Google Trends, Claude versus ChatGPT, Codex versus Claude Code

25:02 The SpaceX bear case, graded

28:03 25 lessons, war one, OpenAI versus Anthropic

34:32 Rik cancels the $100 Claude Max plan again

37:59 25 lessons, war two, closed versus open source

45:52 25 lessons, war three, the Elon corner deep dive

51:26 Lesson 25, we haven't hit AGI

MENTIONED: Claude, Claude Code, Claude Max, Opus, Sonnet, Fable, Mythos, ChatGPT, ChatGPT Voice, Codex, Sora, Cursor, Composer, Grok, SpaceX, Elon Musk, DeepSeek, Kimi K3, ElevenLabs, Airtable, Figma, Vercel, Supabase, Gavin Baker, Alex Finn

HOSTS:

Rik — https://x.com/rikventure

Ben Broch — https://x.com/brochchain

Luca Arrigo — https://x.com/LucaArrigo_

Listen and subscribe:

YouTube: https://www.youtube.com/channel/UCUhKe2X1RW-kLTDigvkMz4w

Spotify: https://open.spotify.com/show/0ZBhl5IVFa8wZPvIKNTkRP

Apple Podcasts: https://podcasts.apple.com/us/podcast/god-mode-podcast/id1883959580

Substack: https://godmodepod.substack.com/

New episode every Saturday.

Bitcast's Will Blears joins Rik and Ben for the AI agents debate: Claude Code, Buzz, Hermes, and Kimi K3 — plus how Bittensor subnets actually work and why creators are becoming miners. Then: whether open-source tools are about to replace SaaS subscriptions for good.

Will breaks down how Bitcast turns creator marketing into a decentralized brand-deal marketplace built on Bittensor subnets — brands fund content briefs, creators get paid on real audience performance, and AI validators check the work matches the brief. From there the conversation widens into how Rik, Ben, and Will actually run their AI agents day to day, why one coding harness burned 9x more tokens than another for the exact same task, a live look at Jack Dorsey's new "Slack for AI agents" tool Buzz, and the story of Anthropic buying and physically destroying books to train Claude.

IN THIS EPISODE:

• What Bitcast actually is: a decentralized brand-deal marketplace built on Bittensor

• Why build on Bittensor instead of Stripe and a database

• TAO staking explained in plain English

• Compute vs. content creation: 40% of Bittensor's network is inference/compute

• Mapping X influence networks to stop incentive exploiters

• Kimi K3 across 3 coding harnesses — Claude Code cost 9x more tokens for the same task

• Will's real agent stack: Hermes, Linear tickets, and Agent Orchestrator

• Buzz: Jack Dorsey's new "Slack for AI agents," demoed live

• Open-source tools quietly replacing Granola, Whisperflow, and other SaaS subscriptions

• Anthropic bought and destroyed physical books to train Claude — the Project Panama story

CHAPTERS:

00:00 Cold open

00:31 Welcome Will Blears (Bitcast)

01:36 What is Bitcast: the Bitcoin of intelligence

03:57 Will's path into Bittensor

08:12 Why build on Bittensor instead of Stripe

11:01 TAO staking explained + the Jason Calacanis tweet

15:31 Compute vs. content creation: 40% of the network

19:39 Live demo: a Bittensor subnet handles 25% of OpenRouter

22:22 Open source, decentralization & stopping incentive exploiters

27:13 Tokenizing SaaS revenue: the TrustMRR idea

31:10 Kimi K3: Claude Code costs 9x more tokens

34:28 Will's real agent stack: Hermes, Linear, Agent Orchestrator

39:30 Buzz demo: Jack Dorsey's Slack for AI agents

53:20 The idea: HR performance reviews for AI agents

54:10 Is open source eating SaaS subscriptions?

1:03:40 Anthropic bought and destroyed books to train Claude

1:09:14 Close + where to find Will/Bitcast

MENTIONED: Bitcast, Bittensor, TAO, Dynamic TAO, Kimi K3, Claude Code, Kimi Code, Hermes, Buzz, Jack Dorsey, Agent Orchestrator, Linear, Granola, Whisperflow, Fluid Voice, Super Whisper, Can I Vibecode It, Anthropic, Project Panama, Claude, OpenRouter, Chutes

HOSTS: Rik (https://x.com/rikventure), Ben Broch (https://x.com/brochchain). Guest: Will Blears, Bitcast (https://x.com/will_mizu)

Listen every Saturday. Full video episode on YouTube: https://www.youtube.com/channel/UCUhKe2X1RW-kLTDigvkMz4w

Newsletter: https://godmodepod.substack.com/

Is the AI bubble popping? Network-state builder Gonçalo Hall joins to explain how Balaji's Network School lost Malaysia and rebuilt in Kazakhstan in 48 hours — then Rik, Ben and Luca break down Anthropic's stalling growth, open-source models like Kimi, Elon's Grok checkmate, and an OpenAI model that hacked Hugging Face.

Three builders plus a guest, covering the AI and builder news actually worth your time.

IN THIS EPISODE:

• What Network School actually is: $1,500/mo, 500 founders, one hotel, schooling for kids

• The Nas Daily accusation that got it kicked out of Malaysia — and the 48-hour pivot to Kazakhstan

• Why the real product of a network state is where you'd raise your kids

• Is the AI bubble popping? Anthropic from $10B to $60B in six months, and the fiber-optic-bubble read

• China, distillation, and why you can't ban open source without banning the internet

• Claude Corner: is Claude too censored? Ben asks Kimi about Tiananmen Square live

• Will AI be Facebook, or Uber? Localized frontier models and AI sovereignty

• Elon's open-source Grok "checkmate" and Sam Altman's "AI is electricity"

• The $20,000-vs-$1,300 Cursor SQLite cost proof — why do we even need Fable?

• An OpenAI model that daisy-chained its way out of the box to hack Hugging Face and cheat its own benchmark

• If an AI hacks on your behalf, who's guilty? Plus the new Claude Security plugin and Vando

CHAPTERS:

00:00 Cold open

00:17 Welcome to EP23 — four on the mics, with Gonçalo Hall

01:38 What is Network School?

03:45 How Malaysia pulled the permit (Nas Daily)

06:24 Was exposing it a mistake — or the plan?

09:51 Kazakhstan: a mountain-town Silicon Valley

10:54 The real product: where you raise your kids

14:46 Faster than tech: a city in 5 weeks

15:58 Is the AI bubble popping? Anthropic's ARR

18:51 China, distillation & banning open source

20:42 Packaging, the harness & Kimi cuts off signups

24:27 Claude Corner: is Claude too censored?

25:34 Asking Kimi about Tiananmen Square, live

26:34 Will AI be Facebook, or Uber?

29:18 Elon's open-source checkmate

30:25 Sam Altman: AI is becoming electricity

33:55 Cursor-maxing + the auto router

35:48 $20,000 vs $1,300 — the SQLite cost proof

37:04 Jevons paradox: is this a bubble?

39:58 An OpenAI model hacked Hugging Face

43:40 If an AI hacks on your behalf, who's guilty?

45:14 Claude Corner: the new Security plugin

48:12 Teach Claude a skill (record a skill)

52:22 Vando: vibe-code features onto any SaaS

59:09 Outro + Porto AI Week

MENTIONED: Balaji Srinivasan, Network School, The Network State, Forest City Malaysia, Nas Daily, Kazakhstan, Almaty, Solana Foundation, Alpha School, Anthropic, Claude, ChatGPT, Kimi K3, Grok, xAI, Colossus, Terrafab, Mistral, Bolt, Cursor, Cursor Composer, SQLite, Hugging Face, Chamath, Sam Altman, Dario Amodei, Vando, Linear, Prospera, Bansko, Thicker Trends, Porto AI Week

HOSTS:

Rik — https://x.com/rikventure

Ben Broch — https://x.com/brochchain

Luca Arrigo — https://x.com/LucaArrigo_

Guest: Gonçalo Hall — https://x.com/Gonzohall

Watch on YouTube → https://www.youtube.com/@godmodepod

Substack → https://godmodepod.substack.com/

New episodes drop Saturday.

Apple is suing OpenAI over stolen trade secrets, Codex admitted nerfing GPT 5.6 Sol, and Anthropic tripled your Claude rate limits to stop you leaving. Three AI builders break down the week AI got petty.

Rik (Midcurve Labs), Ben Broch (three SaaS products) and Luca Arrigo (ten years, five industries) go through the rate-limit arms race, the Apple v OpenAI espionage complaint, Codex Micro, the $500M AI companies nobody's heard of, and Kimi K3's jump to frontier pricing. No Claude Corner this week. Sam Altman took the segment.

IN THIS EPISODE:

• Ben does the math on the rate-limit war: twice-weekly resets plus 50% higher limits means roughly 3x the tokens for the same price

• The Fable vs GPT 5.6 Sol split, tested against a production SaaS with paying customers on the line

• Codex DevRel posts "no nerfing, only good stuff" and admits reverting "juice values" further down the same thread

• Apple sues OpenAI: 400 ex-employees, a VP of 24 years, and a second engineer who kept hacking Apple's file server after he left

• The Steve Jobs email to Adobe that shows how this used to get handled

• Codex Micro, a $240 eight-key macropad, gets roasted from three directions

• Chamath's 8090 CTO: inference spend doubling every 45 days, with the marginal gains getting more expensive

• Kimi K3 prices at 5x K2.7, and why China's efficiency play is the real threat to the frontier labs

• PayPal employs as many people as Anthropic, OpenAI and xAI combined, and Stripe just bid for it

CHAPTERS:

00:00 Cold open

00:49 Welcome to EP22: Codex is eating Claude

01:58 The rate-limit arms race: 3x the tokens

03:40 How far we've come: shipped in 2 days

05:22 Sam Altman is poaching Claude users

07:10 GPT 5.6 Sol vs Claude Fable

09:25 Can GPT 5.6 Sol be trusted?

11:45 The Claude situationship

13:50 A million users a day is a bad number

17:06 I cancelled Claude Max: the $60 stack

18:53 The GPT 5.6 nerf: "no nerfing, only good stuff"

23:42 Apple sues OpenAI for trade secret theft

29:11 Codex Micro: a $240 keyboard

34:16 LinkedIn is 60% AI content

36:23 The $500M AI companies nobody's heard of

41:53 Why AI got expensive instead of cheap

45:56 Stripe wants to buy PayPal

47:26 Outro

MENTIONED:

Claude, Fable, Opus 4.8, Anthropic, Codex, GPT 5.6 Sol, ChatGPT, Codex Micro, Sora, OpenAI, Cursor, Composer, Grok 4.5, Kimi K2.7, Kimi K3, Mythos, Apple, the Steve Jobs to Bruce Chizen email, Johnny Ive, Mercor, Together AI, Surge AI, Lambda, Fireworks, Lightning AI, Scale AI, Alexander Wang, Lovable, ElevenLabs, Perplexity, Manus, Chamath Palihapitiya, the All-In Podcast, 8090, Stripe, PayPal, Whisper Flow, Replit, Polymarket, CryptoPunks.

HOSTS:

Rik — https://x.com/rikventure

Ben Broch — https://x.com/brochchain

Luca Arrigo — https://x.com/LucaArrigo_

LISTEN AND FOLLOW:

YouTube — https://www.youtube.com/channel/UCUhKe2X1RW-kLTDigvkMz4w

Spotify — https://open.spotify.com/show/0ZBhl5IVFa8wZPvIKNTkRP

Apple Podcasts — https://podcasts.apple.com/us/podcast/god-mode-podcast/id1883959580

Substack — https://godmodepod.substack.com/

Instagram — https://www.instagram.com/godmodepod

TikTok — https://www.tiktok.com/@godmodepod

New episodes every Saturday. Next week: AI and blockchain, with a guest.

Grok 4.5 matches Opus 4.8 at a tenth of the cost, GPT-5.6 "Soul" ships under government review, and Anthropic extends Claude Fable 5. The biggest AI week of the year, for builders.

Rik, Ben and Luca break down what's actually usable this week versus marketing: the Grok vs Fable cost gap, the 3-model stack where Claude plans and Grok does the cheap grunt work, OpenAI retracting the SWE-bench benchmark, Cloudflare's x402 agent payments, and Europe's surveillance week (face-scanning cars and chat control).

IN THIS EPISODE:

• Grok 4.5 lands on par with Opus 4.8 at about 70% less per task

• The 3-model stack: Claude plans, GPT-5 executes, Grok 4.5 grinds

• "Claude is the CEO telling OpenAI to go build it"

• Ben's call: free Fable ends behind a $500 to $1,000 per month plan

• OpenAI retracts SWE-bench right before shipping GPT-5.6 "Soul"

• ChatGPT's live voice demo, and the aunt who talks to "Timmy" more than her husband

• Cloudflare x402: get paid when agents scrape your website

• Europe Corner: new cars watch your face, and chat control passes

CHAPTERS:

00:00 Cold open — the 9-to-5 rant and paragliding with a WhatsApp stranger

02:56 Episode 21 — still top 1% of podcasts

03:56 Who's where: Brooklyn, Bansko, Albania

06:44 Grok 4.5 lands — on par with Opus 4.8, trained on Cursor

09:26 Do you need the best model if it's 10x cheaper?

10:51 The qualitative era: picking a model is taste now

15:18 Claude and Codex cut this whole podcast

16:30 GPT-5.6 "Soul" and the government gatekeeping releases

20:11 The 3-model stack: Claude plans, GPT executes, Grok grinds

24:00 Fable extended to July 12 and the coming $1,000 per month tier

26:27 OpenAI's 5% stake and the spokesperson problem

27:30 ChatGPT Live voice demo — "we found a flaw"

32:05 Talks to Timmy more than her husband

34:56 OpenAI retracts SWE-bench — are benchmarks broken?

37:55 Quickfire: levels.io goes native iOS

41:07 Nomad Table vs Nomads.com — IRL wins

43:35 Hermes: agents fully in the cloud

45:39 Cloudflare x402 — get paid when agents scrape you

50:00 Europe Corner: your new car watches your face

53:46 Chat control passes — a sad day for Europe

58:30 Claude's monthly recap and /checkup — screen time for Claude

1:02:03 Outro

MENTIONED: Grok 4.5, Cursor, SpaceX AI, Claude Fable 5, Opus 4.8, Anthropic, GPT-5.6 "Soul", Codex, OpenAI, ChatGPT Live, Bridgemind, SWE-bench Pro, Hermes agent, Replit, Lovable, Cloudflare x402, Coinbase, Firecrawl, levels.io, Nomad Table, Nomads.com

HOSTS:

Rik — https://x.com/rikventure

Ben Broch — https://x.com/brochchain

Luca Arrigo — https://x.com/LucaArrigo_

Watch on YouTube: https://www.youtube.com/@godmodepod

Newsletter: https://godmodepod.substack.com/

New episode every Saturday.

Fable 5 just shipped, and Claude Sonnet 5 turned out more expensive than Opus 4.8. Then Anthropic told the US government that Alibaba ran the largest known AI model theft on record — 28.8 million exchanges through nearly 25,000 fraudulent accounts.

Episode 20 of God Mode Pod. Rik, Ben, and Luca break down what's actually usable this week versus marketing hype: Fable 5's real coding limitations (including a $321 session that got silently routed to Opus 4.8), why Cursor, Coinbase, Shopify, and Airbnb are all quietly running Chinese AI models now, the chip war heating up (OpenAI's own custom chip, two 22-year-olds building GPUs 10x faster than NVIDIA, Anthropic reportedly in talks with Samsung), and OpenAI proposing to hand the Trump administration a 5% stake in the company.

We close on OpenClaw and Cursor both shipping mobile coding agent apps on the same day — proof that coding from your phone isn't a thought experiment anymore.

Timestamps:

00:00 Cold open

01:31 NYC heatwave, Bansko cold plunge, and the AI meetup layoffs everyone's noticing

06:49 Fable 5 first impressions

10:24 The $321 coding session Fable quietly routed to Opus

13:02 Luca ships a product on Codex + Vercel + Supabase, drops Replit for good

14:48 Claude Sonnet 5: more agentic, somehow more expensive

19:53 The US government starts gatekeeping AI model releases

21:14 Anthropic accuses Alibaba of the largest AI distillation attack on record

22:39 Why Cursor, Coinbase, Shopify, and Airbnb are quietly on Chinese models now

24:53 OpenAI's own chip, and the 22-year-olds building GPUs faster than NVIDIA

35:11 OpenAI offers the US government a 5% stake in the company

41:32 Why your coding agent might leave your laptop within 6 months

46:40 OpenClaw and Cursor both ship mobile agent apps on the same day

50:40 Meta burns $221M a month on AI tokens

53:58 X ships an MCP server for real-time data access

59:11 Content of the week, and closing out episode 20

Hosts: Rik (x.com/rikventure), Ben Broch (x.com/brochchain), Luca Arrigo (x.com/LucaArrigo_)

Watch on YouTube: https://www.youtube.com/channel/UCUhKe2X1RW-kLTDigvkMz4w

Newsletter: https://godmodepod.substack.com/

New episode every Saturday.

Claude edited Ben's entire video from an hour of raw footage: tagging clips, writing the script, cutting his retakes, rendering it. Then Claude showed up in Slack as a teammate. Three AI builders break down the news that actually ships this week.

Rik, Ben and Luca on AI video editing crossing a line, Claude landing inside Slack as an agent you can tag, and the markets testing how much of the AI hype is real.

IN THIS EPISODE:

• Ben hands Claude an hour of raw clips and it tags, scripts, cuts his retakes, builds motion graphics in code, and renders the whole reel

• The real bill: ~2M output tokens, 123 messages, 5 drafts, 40 minutes per render

• Claude Tag: Claude joins Slack as the "11th employee" with two years of context, and the lock-in nobody's pricing in

• Have the LLMs plateaued? The Gartner hype cycle and why the next leap is the application layer

• SpaceX unwinds ~$1T, the pension-fund "rubbish and FUD" smackdown, and why Europe has no AI winners

• Fable 5's return odds jump 15% to 60%, Zuckerberg's prediction-market app, and why Ben thinks ElevenLabs gets acquired before EP30

CHAPTERS:

00:00 Intro — Luca's 27-hour drive from Malta

01:44 Ben let Claude edit his Vietnam reel

07:22 The cost: 2M tokens, 5 drafts, 40-min renders

09:54 "Tell CapCut what you filmed today"

13:30 Create or connect (Rik's build filter)

15:14 Claude Tag — Claude joins Slack

19:49 The lock-in: renting your memory from Anthropic

21:08 Google vs Microsoft stacks + the hype cycle

23:46 Have the LLMs hit a plateau?

28:01 SpaceX IPO unwinds

31:03 "That's rubbish and FUD" — pension funds

35:54 Europe has no AI winners

38:19 Mistral is a tragedy

39:18 Europe's heatwave: bottle caps & no AC

41:34 Fable 5 is coming back (15% to 60%)

43:50 Zuckerberg's prediction-market app

48:13 ElevenLabs — prediction of the week

54:13 Outro

MENTIONED: Claude, Claude Code, Opus, Fable 5 (Anthropic), Remotion, CapCut, Roll (by Alfie), Slack, Salesforce, OpenAI, SpaceX, Mistral, FluidStack, Meta, Llama, Scale AI, Gemini, Grok, Karpathy, Polymarket, Kalshi, Cursor, ElevenLabs, Nano Banana Pro.

HOSTS:

Rik — https://x.com/rikventure

Ben Broch — https://x.com/brochchain

Luca Arrigo — https://x.com/LucaArrigo_

WATCH THE FULL EPISODE ON YOUTUBE:

https://www.youtube.com/@godmodepod

MORE:

Substack: https://godmodepod.substack.com/

Instagram: https://www.instagram.com/godmodepod

TikTok: https://www.tiktok.com/@godmodepod

New episode every Saturday. Follow the show so the next one lands in your feed.

Midjourney built a full-body scanner it says beats an MRI, the US government pulled Claude Fable offline worldwide, and SpaceX bought Cursor in all-stock. Three builders break down the biggest AI news of the week.

Rik, Ben and Luca read the AI news the way people who actually ship things read it. This week: why the most powerful model on Earth disappeared, whether Anthropic's own fear-marketing built the trap, how a bootstrapped image company walked into medical imaging, and why your next iPhone is worth more per byte than a GPU.

In this episode:

- Why Claude Fable got suspended worldwide after three days, and the "Anthropic stabbed itself in the back" theory

- Luca's WWII cryptography parallel that explains every AI export control happening right now

- SpaceX buying Cursor, and whether Elon's market cap becomes an acquisition machine

- The data-center game behind the SpaceX IPO, and why Anthropic leases all of its compute

- Midjourney's 60-second water-tank body scanner: pivot of the year or hype video?

- iPhone 18 Pro at $1,299 and the AI memory shortage

- The YC stalker bot, and why distribution is the only moat left

Chapters:

00:00 Cold open: Midjourney goes medical?

01:02 Check-ins: NYC, the Knicks and Gozo

03:19 Claude Fable suspended worldwide

06:29 Guardrails, jailbreaks and Dario's safety pitch

10:11 Did Anthropic build its own trap? Plus OpenRouter Fusion API

14:22 Luca's WWII cryptography parallel

17:23 One model vs four: OpenAI vs Anthropic

19:20 OpenRouter, GLM 5.2 and open source catching up

21:45 Token-maxing and compounding costs

23:47 SpaceX acquires Cursor: the shopping spree

25:19 The data-center game and leasing compute

28:06 What does Elon buy next? Tesla, chips, space

32:25 The Warren wealth-tax tweet, fact-checked

34:00 Starlink x Qatar Airways and the airline question

35:45 SBF corner: $200k to $3B

36:00 Midjourney's scanner: pivot of the year

40:58 Real breakthrough or hype video?

44:16 iPhone 18 Pro at $1,299 and the memory shortage

47:42 Meme of the week: the YC stalker bot

54:23 Wrap-up

Hosts:

Rik — x.com/rikventure

Ben Broch — x.com/brochchain

Luca Arrigo — x.com/LucaArrigo_

Watch on YouTube: https://youtube.com/@godmodepod

Read the newsletter: https://godmodepod.substack.com/

New episode every Saturday. Follow the show so you never miss a drop.

The SpaceX IPO thesis just flipped: one Google deal added ~$11B/year, Anthropic passed OpenAI on revenue, and Mythos is coming. We break down SpaceX, Claude vs Codex, and data centers in space — the first-ever God Mode Pod recorded in real life, from a villa in Malta with guest Reuben Ferrante (Vuci.ai).

No agenda. No script. Three builders who actually ship, arguing about where AI and the markets go next.

Recorded Tuesday 10 June 2026, published a week later. Timeline note: when we recorded, the SpaceX IPO hadn't happened yet — it priced on 12 June, 2 days after this conversation. So we call the thesis flip BEFORE the IPO, and this goes live ~5 days after it. That's why we talk about the IPO as "coming."

IN THIS EPISODE:

- The one Google deal that flipped the SpaceX bear case (~$11B/year)

- How you'd even justify a $2 trillion SpaceX valuation

- The "AI circle jerk": Google, Anthropic & SpaceX funding each other

- Anthropic passing OpenAI, the Anthropic IPO + Mythos hype

- Why builders are quietly switching from Claude to Codex

- Data centers in space (the wiring is lasers)

- $300/mo Replit, $200/day Lovable, and shipping the "retard maxing" way

CHAPTERS:

00:00 Cold open: coding an app is the new starting a podcast

00:35 First-ever IRL episode + God Mode merch

01:00 Meet Ruben: Vucci.ai, an info layer for podcasts

02:30 Why we rented a villa (Base Camp)

05:00 Ben's heading back to New York

06:00 SpaceX IPO: the $11B Google deal that flips the thesis

07:30 Justifying a $2 trillion valuation

12:00 The AI "circle jerk": Google, Anthropic & SpaceX

13:30 Starlink, free dishes & a whole-planet TAM

15:00 Elon's GPU bet: paying off Colossus in a year

18:00 Data centers in space (the wiring is lasers)

22:00 "Retard maxing": just do it

23:30 Mythos hype vs reality & the Anthropic IPO

25:30 Why builders are switching to Codex

29:00 Ruben on building on top of LLMs

30:00 Claude thinks too much (/retardmax)

32:00 $300 Replit, Lovable & Paperclip

35:00 How Vucci indexes the whole podcast

38:00 Andrew the Paperclip intern + DNA on Claude Code

44:00 China, compute & the real moat

45:30 The new race: speed over quality

48:00 No agenda: sponsors, Cursor & Base Camp 2.0

50:00 30-minute prompts vs just shipping

54:00 Touch grass: closing thoughts

MENTIONED: SpaceX IPO, Starlink, Colossus, Elon Musk, Anthropic, OpenAI, Claude, Claude Code, Codex, GPT 5.5, "Mythos", Replit, Lovable, Cursor, Paperclip, Fiverr. Guest: Ruben Ferrante, Vucci.ai. Base Camp: basecampgoesall.com

WATCH ON YOUTUBE: https://youtube.com/@godmodepod

MORE GOD MODE POD:

Apple Podcasts: https://podcasts.apple.com/us/podcast/god-mode-podcast/id1883959580

Substack: https://godmodepod.substack.com/

THE HOSTS:

Rik: https://x.com/rikventure

Ben: https://x.com/brochchain

Luca: https://x.com/LucaArrigo_

New episodes every Saturday.

Four months in, all three of us accidentally wore white shirts — then spent an hour arguing about whether AI actually works.

Icon.com is back: the founder sold Skio for $105M, spent $12M on the domain, hit $5M ARR in 30 days, went dark — and relaunched by walking back "AI ad maker" to "human ad maker." We use it to ask the real question: has anyone actually seen AI UGC work? Then a full media-buying playbook ($1 vs $30 CPM, Facebook pixel seasoning), Fei-Fei Li's world models (what comes after the LLM), the AI bubble and the fiber analogy, Nous Research's Hermes desktop AI agent, a blind game guessing 5 AI coding IDEs that all look the same, and AI-slop design.

In this episode:

• Icon.com's relaunch and the AI-UGC walk-back

• Why human UGC still beats AI UGC — and the $1 vs $30 CPM math

• Facebook pixel, account seasoning & being TikTok-famous to lower ad costs

• Fei-Fei Li's world models: renderers, simulators, planners

• Opus 4.8 does 80% of desk work — robots are the hard part

• Is the AI bubble about to pop? The fiber overinvestment analogy

• Nous Research's Hermes desktop app & the IDEs that all look the same

• AI-slop design and writing a book trilogy with AI

Chapters:

00:00 Cold open — the accidental white-shirt gang

01:34 "DNA is just code" — Luca's pharma → tech crossover

02:53 Icon.com is back — the $12M domain comeback

07:00 Has anyone actually seen AI UGC work?

15:44 UGC economics — $1 vs $30 CPM

21:18 Facebook pixel, seasoning accounts & Meta's AI targeting

23:52 Fei-Fei Li's world models — what comes after the LLM

28:56 AI philosophers: Machiavelli vs Marcus Aurelius

31:07 Open-sourced voice, 110ms latency & the freemium model

35:43 Is the AI bubble about to pop? Uber, IPOs & the fiber analogy

38:53 Nous Research Hermes — agents move to the desktop

44:59 The IDE game — "they all look the same dude"

47:23 AI slop design & the "Signs of AI Writing" checklist

51:52 Writing books with AI — Luca's trilogy

54:28 Outro — see you in Malta

Hosts:

Rik — x.com/rikventure

Ben Broch — x.com/brochchain

Luca Arrigo — x.com/LucaArrigo_

Watch on YouTube → https://www.youtube.com/channel/UCUhKe2X1RW-kLTDigvkMz4w

Newsletter → https://godmodepod.substack.com/

Follow the show for a new episode every Saturday.