Speaker
Ben
Appearances over time
19 episodes
Episodes
19
5 different models dropped last week & the GPT-5.6 usage limits are brutal
We Tested GPT 5.6 Sol Early
Fable Is Back...kinda?
MeidasTouch Full Podcast - 7/2/26
Anthropic Passes OpenAI on Revenue | The SpaceX IPO Thesis Flips | Claude vs Codex | Data Centers in Space | Live From Malta with Ruben Ferrante, Founder @ Vuci.ai | EP17
Does AI UGC Actually Work? | Fei-Fei Li's World Models | Opus 4.8 Can Do 80% of Desk Work | Is the AI Bubble About to Pop? | EP16
Claude Opus 4.8 Drops | Anthropic Rents SpaceX GPUs at $8 an Hour | The SpaceX IPO Bear Case | Token Maxing Is a Vanity Metric | A Browser Inside Codex | EP15
The Bear Case for SpaceX at $1.75T | Karpathy Joins Anthropic | Cursor vs Codex vs Claude Code | Codex Builds an ERP in 7 Hours | Why Ozempic Is $20 in China | EP14
Is Claude Max Worth $100 a Month? | Anthropic Beats OpenAI on Business Adoption | xAI Launches Grok Build | Codex Sells 200 Items in 4 Days | HTML Is the New Markdown | EP13
Codex Has 100x More Downloads Than Claude Code | Anthropic Partners With SpaceX | Anthropic's $380B Compute Stack | Apple iOS 27 Picks Claude and Gemini | EP12
Anthropic Admits Claude Got Dumber | GPT 5.5's 86% Hallucination Rate | SpaceX's $60B Option to Buy Cursor | Naval Launches USVC | A Hairdryer Hacks Polymarket | EP10
Anthropic Nerfed Claude Opus by 73% | Claude Code Hits $2.5B ARR | Anthropic Pre-IPO Shares at $900B | Tesla AI5 Tapes Out | Allbirds Pivots to AI Compute | EP09
Anthropic's Mythos Scores 78 vs GPT-5.4's 58 | Mythos Finds a 27-Year-Old Bug | Anthropic Kills OAuth for OpenClaw | Jensen's $500K Token Rule | Google's Gemma 4 Comeback | EP08
Elon's TerraFab Is 50x TSMC | OpenAI Burned $15M a Day on Sora, Then Killed It | ElevenLabs Undercut 4x Overnight | Anthropic's Mythos Leaks | GitLab Founder Treats His Own Cancer With AI | EP06
The SaaS Apocalypse: Adobe, Figma and IBM | Vibe Medicine: A ChatGPT-Designed Cancer Vaccine for a Dog | Replit $9B vs Cursor $29B vs Lovable $6B | Google Stitch | EP05
AI Can Do 80% of Tasks but We Use 30% | The US Government Runs Last Year's AI | Meta Buys Moltbook | Singapore #1, US #20 in AI Adoption | EP04
Anthropic Missed a $200M Pentagon Deal by 15 Minutes | The Subscription Arbitrage Era Is Over | AI Tokens Got 850x Cheaper | OpenClaw Passes React on GitHub | EP03
Jack Dorsey Cuts Half of Block for AI | White-Collar Jobs Gone in 18 Months? | Why Plumbers Win | Qwen 3.5 and Nano Banana 2 | Running a Business With 10 AI Employees | EP02
OpenAI Acqui-Hires the OpenClaw Creator | OpenAI's $100B Round With Amazon and NVIDIA | Is the AI Bubble Real? | Data Centers in Space | Sam Altman vs Dario Amodei | EP01
Podcasts
Quotes & moments
Anthropic confirmed three separate product-layer bugs ran simultaneously from March 4 to April 16, silently degrading Claude's performance for users.
SpaceX's deal with Google adds approximately $920 million per month — around $11 billion per year — in annual recurring revenue, fundamentally changing the SpaceX valuation thesis.
A developer using AI for coding burns 60 to 100 times more tokens than a casual ChatGPT user, meaning the compute shortage is driven by a tiny fraction of users.
Disney signed a $1B, 3-year deal with OpenAI in December 2025 to use Sora with 200+ licensed characters, then canceled it 90 days later when OpenAI shut Sora down.
On a coding benchmark, Anthropic's Mythos model scored 78 while GPT-5.4 scored 58 — a 20-point gap that suggests a potentially different underlying architecture.
Over the past three months, Claude's search volume grew 120% while ChatGPT's grew only 15%, signaling a major shift toward Anthropic among developers.
The cost of one million AI tokens fell from $60 with GPT-3 DaVinci to 6 cents with open-source Llama 4 — an 850-fold decrease in roughly five years.
Kennen Frost sold his previous company Skio for $105 million in cash before founding Icon.
Figma needed 13 years to reach $1 billion in ARR, while Snowflake needed 9 years — making Claude Code's 9-month trajectory unprecedented.
Ben noted that Block's layoffs represent the largest percentage-based job cut of any S&P 500 company ever recorded.
Claude Code's default reasoning was quietly downgraded from 'high' to 'medium' for an entire month without users being informed.
One user was reported to have consumed 865 million tokens in a single month, costing approximately $2,500 in API fees.
Ben pays $200/month for a Claude Max subscription but consumes roughly 500 million tokens — equivalent to $12,500 at API pricing — demonstrating how broken subscription economics are.
Kennen Frost reportedly spent $12 million to acquire the domain icon.com.
Roughly 5% of top Anthropic users account for approximately 80% of all token usage on the platform.
Icon tried to be 100% AI from start to finish, and it failed. The winning formula is 80% AI with a human in the loop for taste, editing, and final execution. The human layer isn't a weakness — it's what makes the product credible.
A viral UGC post can deliver $1 CPM. The same content boosted through Meta paid ads in New York costs $30 CPM. That's a 30x difference — and no AI platform has closed that gap.
Icon.com swapped one phrase on its website — 'AI ad maker' to 'human ad maker' — and that single change reveals the state of AI video advertising. AI UGC doesn't work yet, and the platforms betting on it are quietly admitting defeat.
Before you run a single paid ad, you need at least 1,000 users cookied on your Meta pixel. Luca was shocked to meet AI builders who had never heard of pixel seasoning — the foundational move that unlocks retargeting and lookalike audiences.
The 1990s fiber build-out overran by years and crushed investor returns — even though it eventually worked. The AI compute boom may be heading the same way, with companies like Uber and Microsoft already cutting AI spend just as Anthropic and OpenAI eye public markets.
Hermes just launched a desktop app that looks identical to Claude, Codex, and Cursor — because they all do. The real unlock is that you can swap any model into the backend. Vendor lock-in in AI agents is ending.
There's a Wikipedia article on signs of AI writing, and Ben had all his apps rewritten to remove the traces. The four visual tells of vibe-coded apps are becoming a liability — Google and Bing are likely already deprioritising them.
Natural conversation needs a response time under 200 milliseconds. A new open-sourced voice model clocks in at 110ms — one tenth of a second. That's not just fast, it's the threshold that makes AI voice feel genuinely human.
In 2016, Luca wrote a movie plot in his journal about a million-person social experiment that proves even a perfectly balanced society still finds reasons to discriminate. Now he's using AI to write it, trailer it, and potentially turn it into a TV series.
Hermes lets you run a self-improving agent on a VPS over Telegram or a desktop app. OpenAI and Anthropic keep you locked to user-facing apps. Ben wonders: is that a safety decision, a business model decision, or just a failure of ambition?
Go viral on TikTok or X, pixel every visitor, then retarget that warm audience through Meta's precision targeting. Luca's version: dance on TikTok until 30% of Malta knows your face, then run ads to people who already recognise you. CPM drops accordingly.
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.
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.
LLMs only capture the statistical structure of text. Fei-Fei Li's world model taxonomy adds renderers, simulators, and planners — and the simulation layer is the underrated linchpin that could unlock true AGI and slash the cost of AI video.
Analysis
What they talk about
- Technology 57%
- Business 29%
- Education 6%
- Government 6%
- Health & Fitness 1%
- Society & Culture 1%
Connections
Shows they appear on and people they share episodes with. Drag to explore.
