Search for a command to run...

Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
In this first episode of 2026, Jason Calacanis and Alex Wilhelm dive into major startup and tech developments from the holiday break, including NVIDIA's massive $20 billion acquisition-hire of Groq's inference technology and the growing impact of AI on traditional businesses. (06:41) They explore how GLP-1 medications are disrupting the restaurant industry, discuss the drama surrounding Yann LeCun's departure from Meta to start his own AI company, and examine prediction markets as a new asset class. The hosts also analyze emerging trends in legal tech, where tools like NotebookLM are collapsing thousands of dollars in legal work into simple AI queries. (46:19)
Jason Calacanis is a renowned angel investor, entrepreneur, and host of This Week in Startups. He's an early investor in companies like Uber, has launched multiple successful ventures including the Launch Accelerator, and founded Founder University with programs expanding globally to Japan and Saudi Arabia. He's also the founder of Angel.co (formerly syndicate.com) and has invested in hundreds of startups through his various funds.
Alex Wilhelm is a technology journalist and co-host of This Week in Startups. He writes the newsletter "Cautious Optimism" at cautiousoptimism.news and has extensive experience covering venture capital, startup funding, and technology trends. He brings a data-driven perspective to startup analysis and market dynamics.
Fred Wilson's experiment with NotebookLM demonstrates how AI can collapse $50,000 worth of legal due diligence into a few hours of work using free tools. (46:19) By uploading prior deals and startup legal documents into separate notebooks, Wilson was able to identify contract differences and legal issues that would typically require extensive lawyer review. This represents a fundamental shift where founders can handle more legal work independently, similar to how cloud computing eliminated the need for expensive server infrastructure. The implications are profound: more startups will be able to bootstrap longer, legal costs will decrease industry-wide, and there will be 10x more startups as barriers to entry lower.
NVIDIA's $20 billion licensing deal with Groq represents a new model for major acquisitions that bypasses traditional regulatory hurdles. (06:41) Rather than acquiring the entire company, NVIDIA licensed Groq's inference technology non-exclusively, allowing for immediate deal closure without FTC approval delays. This structure enables faster transactions but comes with different tax implications - likely ordinary income rather than capital gains treatment. For founders, this creates new exit opportunities and demonstrates that companies with market caps exceeding $3 trillion should be acquiring 20+ companies annually to maximize their high valuations.
A world-famous chef told Jason that the restaurant business is "over" due to the combined impact of GLP-1 medications (like Ozempic), reduced alcohol consumption, and changing social behaviors among younger demographics. (02:16) Customers now order half as much food, spend around $25 per person instead of $75, and prefer cannabis or abstinence over alcohol. This trend is eliminating mid-tier restaurants while preserving only high-end ($200-300 per person) and fast-casual options like Sweetgreen. The broader lesson: when societal behaviors shift due to technology or medications, entire industries can be disrupted, creating opportunities for new business models.
Yann LeCun's departure from Meta highlights a fundamental philosophical divide in AI development between Large Language Models (guess-the-next-word) and world modeling approaches. (21:52) LeCun believes that understanding physical reality - like recognizing gondola cables to avoid fatal accidents - requires more than literary pattern matching. This represents the next frontier where AI must comprehend 3D environments, physics, and real-world consequences. For entrepreneurs, this suggests opportunities in robotics, autonomous vehicles, and any application requiring AI to interact with physical spaces rather than just text.
Polymarket and similar platforms are becoming sophisticated tools for startup intelligence, with active markets predicting which companies will IPO in 2026. (35:58) Jason identified several mispriced opportunities: SpaceX (74% seems low, should be 90%), Anthropic (50% seems low, should be 80%), and Anduril (49% seems low, should be 70%). These markets aggregate information from thousands of participants and can provide valuable insights for investors, founders, and analysts tracking startup trajectories. The key is recognizing when markets are mispricing based on insider knowledge of industry dynamics.