| Episode | Status |
|---|---|
| Episode | Status |
|---|---|
This week on Built 2 Scale, Matt navigates Austin life while Scotty recovers from his US adventure. We break down the massive AI shakeups: GPT-5's controversial launch causing user rebellion, Meta's A...
This episode covers major AI industry shakeups including Meta's AI division downsizing with Yann LeCun on the chopping block, GPT-5's controversial router system causing user backlash, and NVIDIA's bold move to compete directly with clients by building robots. The hosts dive deep into AI infrastructure spending, agent-building tools like Lindy.ai, and the physical AI boom with companies like Luminar Vehicles building autonomous excavators in just two years. Key themes include the value capture problem in AI, founder-led velocity versus corporate inertia, and practical applications of AI agents in business workflows.
Meta officially announces downsizing of its AI division with executives expected to leave. Yann LeCun is on day 19 of a 35-day 'lettuce watch' prediction, with Alexander Wang's recent hire as CTO increasing pressure. The hosts predict LeCun's departure is imminent, citing his public fights with prominent figures rather than focusing on his role.
OpenAI's GPT-5 release created unexpected backlash as users revolt against the new routing system that automatically selects models. The router merges multiple models (O3, 4.1, 4.5) but struggles to make good decisions, forcing users to add more detailed prompts. This represents a strategic shift toward mass market distribution over power users, with free users in India becoming the largest market.
8 Sleep raised $100M to build what they call the 'operating system of sleep and longevity' using 1 billion hours of biometric sleep data. Their strategy exemplifies the new AI playbook: collect proprietary data, create AI agents that optimize outcomes, and become the ecosystem. The discussion explores whether this repeatable pattern works better for horizontal B2C plays versus vertical B2B solutions.
Semi Analysis founder Dylan Patel reveals the broken value capture in AI: OpenAI creates 10-20% of value but captures only 1%, while NVIDIA captures 80% of profits. He advises OpenAI to become a payments platform, taking a slice of every agent transaction. The discussion covers massive infrastructure spending: Meta at $17B quarterly CapEx, Microsoft planning $30B, and Google targeting $85B for 2025.
Scotty completes Lindy.ai's one-week academy, building three agents including one that automatically researches leads, creates psychological profiles, and prepares meeting cheat sheets. The discussion breaks down the evolution from workflows to AI-enabled workflows to true agents that make autonomous decisions. Key insight: agents need explicit instructions to not ask questions and keep working until completion.
Analysis of why founder-led companies (OpenAI, Grok, Claude, Shopify, NVIDIA) dominate AI innovation while corporate CEOs at Microsoft and Apple struggle despite massive resources. Google's turnaround after Larry and Sergei returned demonstrates the pattern. Founders prioritize legacy over quarterly earnings and have structural advantages like supervoting shares, enabling bigger risks and faster pivots.
NVIDIA announces partnership with Foxconn to build robots, showcasing them in November 2024 as the 'beginning of the physical AI era.' This represents vertical integration from chips to hardware to software, directly competing with clients like Tesla, OpenAI, and Google. Jensen is leveraging his dominant position to move up the stack into applications, similar to how OpenAI is eating app-layer companies.
Luminar Vehicles CEO Ahmed Shuber (in his 20s) shares journey from broken lawnmower prototype in 2022 to near-production autonomous bulldozers and excavators in 2024. The story exemplifies extreme velocity in hardware, challenging assumptions about what's possible. Discussion explores why established players like Caterpillar (founded 1925) can't match startup speed despite resources.
Shopify CEO Toby Lütke advocates for capturing all data but not being led by it, arguing that every KPI can be gamed and over-optimization on metrics leads away from core mission. He promotes 80% intuition/taste/emotion with 20% data-driven decision making. Example: Amazon's support metrics showed 1-minute response times, but Jeff Bezos calling support line revealed 15-minute waits.
NVIDIA's Trojan horse, Yann LeCun Timebomb, 8 Sleep Mega Raise and the Physical AI Boom
Ask me anything about this podcast episode...
Try asking: