Why Amazon Will Fight AI Agents to the Death
Amazon makes over $70B in high-margin ads. John Coogan explains why AI checkout agents threaten their entire profit model.
Harvey’s Margins, Are Insects Worth More Than Humans?, Meta Tests Human Help for Muse | Diet TBPN
Hosts John Coogan and Jordi Hays discuss the unit economics of legal AI giant Harvey following reports of negative gross margins driven by agentic token usage. They also examine the viral Effective Altruism debate around insect suffering, Meta's strategy of using human contractors behind its Muse AI agent, and the looming e-commerce battle between Amazon and third-party AI agents.
Key takeaways
- Amazon generates over $70 billion in high-margin advertising revenue, while its core e-commerce net income sits around $36 billion. The retail business does not turn a profit without ad money. Read more →
- An Effective Altruist writer named Bentham's Bulldog argued that aggregate insect suffering outweighs all human suffering combined. Read more →
- Harvey saw its gross margin collapse from positive 50% to negative 50% in June 2026 after agent token usage jumped 20-fold on frontier foundation models. Read more →
- Bloomberg reported that Meta is quietly deploying human contractors to handle live phone calls and concierge tasks for its upcoming Muse personal AI assistant. Read more →
- Grok 4.7 launched with benchmark results targeted specifically at legal tasks evaluated by cost efficiency. Read more →
5 articles from this episode
Why Effective Altruists Care About Insect Suffering
TBPN unpacks the viral EA debate on insect suffering and why aggregate utilitarian math is really about future AI alignment.
Why Agentic AI Broke Harvey's Gross Margins
How Harvey dropped to -50% gross margins when agent token use jumped 20x, and how routing open-weight models fixed their legal tech economics.
Why Meta Put Humans Behind Its Muse AI Agent
TBPN hosts John Coogan and Jordi Hays debate Meta using human contractors for Muse AI calls, revealing the real reason behind human fallbacks.
Grok 4.7 and Opus 5.5 Prove the Model Wars Are Now About Cost
Grok 4.7, Opus 5.5, and GPT-6 Soul shift AI competition toward cheaper tokens, domain benchmarks, and direct margin relief for agent startups.
More TBPN episodes
- Jensen’s Ezra Interview, EV News, Nathan Fielder’s Elizabeth Holmes Doc | Diet TBPNSep 27 · 3 articles
- John Arnold Joins, 4 Tech Arguments (Number 3 Will Shock You)Sep 27 · 6 articles
- Meta Connect 2026, OpenAI’s Australia ‘Incident,’ Bentley’s New EV | Diet TBPNSep 27 · 4 articles
- Meta Connect Reactions, Zuck's Beer Pong Controversy, ClusterMAX 3.0, New Bentley EVSep 27 · 5 articles