Chinese coffee hits NYC, RIP Hulu, GPT-5 arrives and more | E2161

This Week in StartupsAbout 7 min readAug 8, 2025Watch original
THE SUMMARYAI-generated

Key Concepts

  • Luckin Coffee: Chinese coffee chain expanding globally, known for low prices and mobile ordering.
  • Price Dumping: Selling products at unfairly low prices to undercut competitors.
  • Open Weights vs. Open Source: Open weights allow modification of model parameters but not access to the source code, unlike open source.
  • AI Model Training: The process of feeding data to an AI to learn patterns and relationships.
  • User Agent: A string of text that identifies the browser and operating system being used to access a website.
  • Web Scraping: Automatically extracting data from websites.
  • Paper Crawl: Cloudflare's system for allowing AI models to crawl websites for a fee.
  • Right to Publicity: The right of a person to control the commercial use of their name, image, or likeness.
  • AI Hallucinations: Instances where AI models generate incorrect or nonsensical information.

Luckin Coffee's US Expansion and Controversy

Luckin Coffee, a Chinese coffee chain, is expanding into the US market, specifically New York City, after emerging from bankruptcy following an accounting fraud scandal where they admitted to falsifying $310 million in revenue in 2019. Despite the fraud, Luckin has seen significant growth, with 26,026 locations worldwide and a 47% year-over-year revenue increase in Q2 2025, reaching $1.72 billion.

Key Points:

  • Business Model: Luckin focuses on mobile ordering and pickup, enabling higher efficiency and lower prices compared to traditional coffee shops like Starbucks.
  • Pricing Strategy: Luckin leverages inexpensive ingredients (sugar, milk, ice) and a variety of boba-like drinks to offer lower prices.
  • Controversy: The company's Chinese origins and connections to the Chinese Communist Party (CCP) have sparked some backlash, particularly on TikTok, with concerns about the CCP "owning" New York City through coffee.
  • "Communist Tsunami" Argument: Jason Calacanis argues that Luckin's expansion is part of a deliberate strategy by China to undercut American brands, similar to BYD in the automotive industry and TikTok in social media. He suggests imposing tariffs and requiring local partnerships with American involvement.
  • Counter-Argument: Alex Wilhelm argues that restricting consumer access to lower-priced coffee reduces consumer welfare and that demanding "golden shares" in Luckin USA would mirror CCP practices. He suggests that the real issue is price dumping, which should be addressed.

Example:

  • A TikTok video features a woman in a hammer and sickle jersey visiting a Luckin Coffee store, praising its low prices and criticizing capitalism. She notes that Luckin went through a $300 million accounting scandal where they faked their profits because capitalism sucks. And then they started paying their workers and putting the money back into the coffee and what actually matters. And now they are more successful than Starbucks.

Disney Plus and the Streaming Landscape

Disney is consolidating its streaming services by integrating Hulu into Disney Plus, aiming for a unified platform by 2026. This move is expected to improve user experience, reduce churn, and streamline ad sales.

Key Points:

  • Integration: Hulu will be sunset as a standalone app and platform, becoming a tab within Disney Plus.
  • Rationale: Bob Iger believes this will create a cleaner user experience and reduce subscriber churn.
  • Subscriber Base: As of the end of last year, Disney Plus had approximately 128 million subscribers, while Hulu had an additional 52 million. Netflix has just over 300 million global subscribers.
  • Historical Context: Jason Calacanis recalls a prediction he made on CNBC in 2017, stating that Disney would eventually compete with and potentially surpass Netflix in the streaming market.
  • Merchandising and Parks Integration: Jason Calacanis suggests integrating Disney merchandise and park access into the Disney Plus app, offering exclusive benefits to subscribers, such as early access to park tickets and discounts on merchandise.

Examples:

  • Jason Calacanis suggests that Disney Plus should have a tab for Disney merchandise, offering discounts and exclusive items to subscribers based on their viewing history.
  • He also proposes giving Disney Plus subscribers priority access to Disney park tickets, especially during peak seasons like holidays.

AI-Generated Content and the "Adam" Controversy

Zach Dives, a YC Winter 25 founder, created a marketing campaign for his AI-powered CAD editor and 3D modeling tool, Adam, that involved a fake "All-In" podcast segment and a TechCrunch article, suggesting that the AI was actually a human named Adam. This caused confusion and led to Jason Calacanis asking Zach to remove his likeness from the campaign.

Key Points:

  • Marketing Tactic: Zach used AI tools to create realistic deepfakes of Jason Calacanis and a TechCrunch article to promote Adam.
  • Confusion and Backlash: The campaign was so convincing that it confused some viewers, leading to concerns about misinformation and the misuse of AI.
  • Right to Publicity: Jason Calacanis emphasized the importance of obtaining permission before using someone's likeness in advertising, even in parody.
  • Deepfake Crypto Ads: Jason Calacanis expressed concern about the proliferation of deepfake crypto ads featuring celebrities, which are used to scam viewers.
  • Adam's Functionality: Adam is an AI-powered CAD editor and 3D modeling tool that allows users to create 3D models by simply describing what they want.

Examples:

  • Zach created a deepfake video of Jason Calacanis discussing Adam on a fake "All-In" podcast segment.
  • He also created a Muppet version of Jason Calacanis to further promote the company.

Open Weights AI Models and Privacy Concerns

OpenAI released two "open weights" AI models, allowing users to download and run them locally. This is a step towards more transparency but differs from open source, as the source code is not accessible.

Key Points:

  • Open Weights vs. Open Source: Open weights allow modification of model parameters but not access to the source code, unlike open source.
  • Local Execution: Users can run these models on their own computers, eliminating the need to rely on cloud-based APIs.
  • Privacy Implications: Running AI models locally raises privacy concerns, as users may be processing sensitive data (e.g., personal photos, legal documents) without sending it to a third-party server.
  • Apple's Potential Role: Jason Calacanis predicts that Apple will integrate encrypted AI models into its operating systems, allowing users to perform AI tasks locally without compromising their privacy.
  • Nvidia's Stance: Nvidia has stated that its hardware does not contain kill switches or backdoors, addressing concerns about potential government access to user data.

Examples:

  • Users can use these models to create videos from personal photos without uploading them to a cloud service.
  • Businesses can use these models to analyze sensitive legal documents without exposing them to external servers.

Cloudflare vs. Perplexity and the Ethics of Web Scraping

Cloudflare accused Perplexity, an AI-powered search engine, of unethical web scraping practices. Cloudflare claimed that Perplexity was circumventing website restrictions and accessing content without permission.

Key Points:

  • Paper Crawl: Cloudflare has developed a system called Paper Crawl, which allows AI models to crawl websites for a fee, providing a mechanism for content creators to monetize their data.
  • Cloudflare's Accusation: Cloudflare accused Perplexity of bypassing website restrictions and scraping content without permission.
  • Perplexity's Response: Perplexity denied the accusations, claiming that it respects website restrictions and that Cloudflare misunderstood the nature of its user agent requests.
  • User Agent Confusion: The dispute may stem from a misunderstanding of how Perplexity's browser extension, Comet, interacts with websites. Comet users may be inadvertently triggering web scraping activity without their knowledge.
  • Automation and Scale: The core issue is the automation and scale at which AI models can access and process information, raising concerns about the impact on content creators and website owners.

Examples:

  • Cloudflare set up a test website with a "stop sign" for AI crawlers and found that Perplexity's models accessed the site despite the restriction.
  • Jason Calacanis uses Comet to summarize Slack channels, which may be interpreted as web scraping activity by website owners.

Additional AI Developments

  • DeepMind's Genie 3: A world model with improved context maintenance, allowing for more realistic and immersive AI experiences.
  • Anthropic's Claude 4.1: An incremental improvement to the Claude 4 language model, offering better performance and capabilities.
  • GPT-5 Release: OpenAI is expected to release GPT-5, with improvements in coding and reasoning abilities.

Conclusion

The episode covers a range of topics, from the global expansion of a controversial coffee chain to the ethical implications of AI-generated content and web scraping. It highlights the challenges and opportunities presented by emerging technologies, emphasizing the importance of responsible innovation and ethical considerations. The discussion also touches on the evolving landscape of the streaming industry and the ongoing competition between major players like Disney and Netflix.

AI summaries can miss context or contain errors. Check important details against the original video.

MAKE IT YOURS

Read. Remember. Reuse.

Free tools

Go a little deeper.

Have a question about this video? Load its transcript to open the video chat.