Harvey’s Margins, Are Insects Worth More Than Humans?, Meta Tests Human Help for Muse | Diet TBPN

Overview

This episode dives into the challenges faced by Harvey, a legal AI company, as it navigates financial pressures and competitive dynamics in the AI space. The hosts also explore philosophical debates about the moral worth of insects versus humans, Meta's experiments with human-assisted AI agents, and the ongoing evolution of AI models in the tech industry.

Notable Quotes

- The hardest thing about building Harvey is doing what's best for our customers despite immense pressure to do what's easy. – Gabe, Harvey co-founder, on prioritizing customer trust over short-term margins.

- If you truly believe that a bug's life is worth more than a human life, you will immediately feed yourself to bugs. Do it right now. – Dumer, reacting to the viral essay on insect welfare.

- Amazon is like the strongest in the AI era of the hyperscalers because they own logistics, infrastructure, the final step, the real world. – John Coogan, on Amazon's dominance in AI-driven commerce.

🧢 Harvey’s Financial Struggles and Strategic Choices

- Harvey faced negative gross margins (-50%) in June due to skyrocketing token usage, but quickly rebounded to positive margins by optimizing model usage and fine-tuning their own AI, Harvey Tenet.

- John Coogan highlighted the competitive pressures in the legal AI space, noting that rivals like Legora question the value of training proprietary models when public models are improving and becoming cheaper.

- Harvey's co-founder, Gabe, defended their approach, emphasizing customer-centric decisions like delaying consumption-based pricing and investing in infrastructure to help clients manage costs.

🐜 Insects vs. Humans: A Philosophical Debate

- A viral essay argued that the collective suffering of insects outweighs human suffering due to their sheer numbers, sparking heated reactions online.

- John Coogan and Jordi Hays dissected the argument, noting its implications for future AI ethics, where humans might be treated as insects by superintelligent machines.

- Critics mocked the essay's premise, with one commenter suggesting that true believers should sacrifice themselves to bugs to prove their commitment.

🤖 Meta’s Human-Assisted AI Agents

- Meta is testing human concierges for its AI assistant, Muse, to handle tasks beyond current model capabilities, such as complex phone calls.

- John Coogan speculated that this approach doubles as a way to collect training data for future model improvements, rather than being a long-term solution.

- The hosts debated the sustainability of human-in-the-loop systems, given advancements in voice models and the scale of Meta's user base.

📈 AI Model Wars and Frontier Pacing

- New AI models like Opus 5.5 and GPT-6 Soul were launched, emphasizing safety and cost-efficiency rather than pushing superhuman capabilities.

- John Coogan noted that Harvey's legal benchmarks influenced Grok 4.7's design, making it particularly cost-effective for legal tasks.

- The hosts discussed the broader implications of pacing the frontier in AI development, balancing innovation with safety.

🛒 Amazon’s AI Commerce Strategy

- Amazon's refusal to integrate directly with OpenAI's instant checkout process highlights its focus on protecting its $70 billion ad revenue stream.

- John Coogan explained how Amazon's dominance in logistics and infrastructure positions it as a key player in the AI-driven commerce landscape.

- Shopify's partnership with Muse was also discussed, with concerns about reduced conversion rates when shopping via AI agents compared to native apps.