Speaker
Rik
Appearances over time
10 episodes
Episodes
10
EP25: 25 lessons in 25 episodes
EP21: The new AI stack — Claude Fable plans, GPT5.6 builds, Grok 4.5 grinds
Claude Opus 4.8, browser in Codex, tokenmaxxing, and the upcoming AI IPO bear case nobody wants to hear
SpaceX's $2T IPO, Karpathy Joins Anthropic & the SaaSpocalypse — ft. Dr Dan | EP14
Anthropic won the AI race, OpenAI gives Codex for free, xAI Grok Build & Claude cancel culture | GMP EP13
Codex beats Claude Code, Stripe launched the AI bank & SpaceX targets a $7.5T valuation | GMP EP11
Anthropic made Opus dumber (73% Drop), Claude Code Hits $2.5B ARR & Tesla AI5 | God Mode Pod EP09
Anthropic's Mythos Beat GPT-5.4, Found 27-Year Bug & NVIDIA's $500K Rule | GodModePod EP08
AI Cures Cancer, OpenAI Kills Sora & Anthropic's Secret Model Leaked | God Mode Pod Ep. 6
``` OpenAI's Power Grab, Sam vs Dario & Why AI Is Turning Everyone Into a Programmer | GMP EP01
Podcasts
Quotes & moments
Grok 4.5 performs on par with Claude Opus 4.8 on key benchmarks but costs approximately 70% less per task, making it compelling for high-volume agentic workloads.
A developer tested Claude Code and Codex subscriptions for 30 days and unlocked $50,000 in API value for only $800 in subscription costs.
SpaceX filed for its IPO targeting a $1.75 trillion valuation, raising $75 billion — making it the largest IPO in history.
Claude Code reached $2.5 billion in annualized recurring revenue just nine months after launch, making it the fastest B2B SaaS product to reach that milestone ever.
God Mode Pod hit one million total views by episode 15.
GPT 5.5 reportedly performs better inside Cursor than inside OpenAI's own Codex environment.
Cursor reportedly hit a $3 billion annual revenue run rate ahead of the SpaceX acquisition deal, though Dan suggested it likely costs $6 billion to generate that revenue.
OpenAI is finalizing a historic $100 billion funding round, with NVIDIA allegedly contributing $30 billion.
Grok 4.5 jumped from 39% to 81% on TerminalBench compared to Grok 4.3, representing a dramatic improvement in agentic coding capability.
The $400 Claude Max plan delivered $32,000 in API equivalent value, an 81x return over the subscription price.
Opus 5 can spend nearly 100 million tokens on a task where a cheaper model like Terra uses only 1.5 million, a ~65x difference in token consumption.
On Artificial Analysis benchmarks, Grok jumped from approximately 1,000 points (Grok 4.3) to 1,500 points (Grok 4.5), landing just behind Claude Opus and Sonnet.
Rik cancelled his $100/month Claude Max subscription live on air, reallocating the spend across Cursor ($20), Codex ($20), Claude ($20), and xAI ($20), with $20 left for video generation.
Per Ramp data, Anthropic overtook OpenAI in business adoption for the first time, despite OpenAI holding roughly 85% of consumer user share versus Anthropic's 5–10%.
Claude Opus 4.8 reduced its dishonesty rate from 19.7% in Opus 4.7 to 3.7%, the biggest measurable improvement in the release.
Token maxing is the new 9-to-9 office presence — it looks productive but proves nothing. Luca argues that measuring success by tokens spent is as meaningless as judging workers by hours logged, while Ben fires back that you can't find the 20% that works without spending through the other 80%.
H200 GPU spot rental prices have fallen from a peak of $7/hr to $4.11/hr, coinciding almost exactly with Anthropic's SpaceX compute deal removing a major source of spot demand. Commoditization of compute is now live and tradeable on Polymarket prediction markets.
Opus 4.8 is less a technical breakthrough and more a press release timed to Anthropic's fundraise at a valuation higher than OpenAI. The main measurable win: dishonesty rate dropped from 19.7% to 3.7%. Everything else feels like post-training polish, not pre-training reinvention.
The next paradigm isn't using AI alongside your browser — it's your browser living inside your AI agent. Log into Supabase once and every future project can use it automatically. Your agent watches you work in Notion, answers questions about what you've done before, and turns one-time workflows into repeatable automations.
Three 2026 AI IPOs plan to raise $550B — more than all US IPOs from 2015 to 2025 combined. SpaceX enters at 100x revenue (NVIDIA trades at 25x), claims a TAM larger than China's GDP, and faces looming insider unlocks starting day 45. The emotional FOMO is real, but so is the math.
Anthropic's $1.25B/month deal with SpaceX for ~200,000 GPUs works out to $8 per GPU per hour — about double the current spot market price. Elon negotiated the better end of that trade, and Anthropic now has the compute firepower to run 1,000-agent swarms that will burn through tokens at unprecedented speed.
App onboarding forms are dead. The future: copy a prompt into Claude, paste the agent's response back to the app, and it configures itself around you. No clicks, no checkboxes, no 'where did you hear about us?' — just personalization at agent speed.
SpaceX gives retail 30% of its IPO shares (6x the usual allocation) while floating only 5% on day one. New regulations push institutional money in within 15 days. Result: a short window where buying pressure overwhelms supply — then insider unlocks start on day 45 and the selling begins.
DNA sequencing spits out a messy FASTQ file. AI cleans it into a BAM file with quality scores. You then compare it to a reference genome to get a VCF of your unique variants. That's your genetic fingerprint — and every step of the pipeline is now an AI problem. Luca is already testing this on open-source DNA.
The dead internet theory is no longer a theory. AI-generated content has so thoroughly contaminated the web that developers are buying pre-2021 human-written code to validate training data. If it existed before 2021, there's a chance it was real. After that — who knows.
SpaceX's S-1 doesn't sell a rocket company. It sells the world's biggest AI infrastructure story. The $28.5T TAM they claim is mostly AI and SaaS services, not launch contracts. The space business is only $5.5T of that — the rest is essentially competing with AWS.
Amazon reportedly spent $500 million in a single month on AI tokens after employees had no usage limits on their cloud accounts. If you're setting AI strategy by how many tokens you burn, this is where that logic ends — a nine-figure cloud bill, no cap, no return measurement.
Coding subscriptions from Anthropic and OpenAI are deliberately priced below cost to capture market share. APIs carry the actual margins. Cursor is doing the same thing — just burning SpaceX's money instead of Microsoft's.
Cursor fine-tunes Chinese models, has no frontier AI, and its IDE is just a VS Code fork. The only real assets are user data and brand awareness — and Elon is burning SpaceX money to buy both. Dan gives seven to eight times more value per dollar from Codex than Claude Code.
The real optimization isn't frontier vs. budget models — it's knowing which task needs PhD-level reasoning vs. fast-enough. Luca uses Gemini and Grok for quick research, Opus for deep analysis; Ben stays all-in on Claude at $200/month; Rik is splitting between Claude Code and Codex, considering dropping to $20 Claude.
Analysis
What they talk about
- Technology 56%
- Business 41%
- Society & Culture 3%
Connections
Shows they appear on and people they share episodes with. Drag to explore.
