Descriptions:
The 20VC podcast with Harry Stebbings, Jason Calacanis, and Rory covers three major business stories in a single episode. The lead item is Anthropic’s $45 billion fundraise from hyperscalers, examined alongside reported compute shortfalls and the concurrent White House restriction on Claude Mythos access. The panel debates whether Anthropic’s 2025 growth reflected genuine product superiority or temporary market share capture during a period when OpenAI failed to ship competitive models — and whether that advantage survives as GPT-5.5 closes the gap in coding performance.
A central forward-looking thesis is that agents — not human developers — will increasingly be the entities choosing which LLMs to use for any given task. If true, this erodes the developer-preference advantage that Claude and Claude Code built last year, since agents may optimize on different signals than humans do. The panel treats this structural shift as one of the most consequential questions for how the LLM market gets monetized over the next several years, with one panelist noting he already sees equal agent performance from OpenAI models in his own workflows.
The episode also covers China’s decision to block Meta’s $2 billion acquisition of Manis, with analysis of where China’s leverage actually sits — over personnel still in country and over companies with significant Chinese operations, not over venture investors who have already received capital distributions. The third story is Thoma Bravo handing Medallia’s keys to creditors and wiping out $5.1 billion in equity, used as a case study on over-leveraged pre-AI software buyouts that failed to transition their business model before debt obligations became unsustainable.
📺 Source: 20VC with Harry Stebbings · Published April 30, 2026
🏷️ Format: Podcast
