The Twenty Minute VCOpenAI Buys TBPN & Their Management Team Reboot | Mercor Hack & Why Now is the Time for Cyber
CHAPTERS
Anthropic’s revenue surge: $30B run-rate and what it signals
The hosts react to reports that Anthropic has exploded from roughly $9B to $30B in a short span, overtaking OpenAI. They frame it as a rare moment where a top competitor is both growing faster and appearing structurally advantaged.
Compute rationing and pricing power: tokens, agents, and capacity allocation
They dig into what happens when a model provider is compute constrained: you ration access, price discriminate, and de-emphasize low-ROI workloads. Anthropic’s moves around agent usage and token-heavy features are discussed as rational economics under scarcity.
Cost structure shock: Anthropic training costs vs OpenAI and investor implications
A leaked comparison that Anthropic’s training costs may be ~¼ of OpenAI’s becomes a focal point. They argue that a cost advantage plus faster growth compounds into a serious strategic and valuation problem for OpenAI.
OpenAI’s last round dissected: tranche financing, compute offsets, and “barely real” capital
They debate how much of OpenAI’s financing was truly cash versus conditional tranches, IPO/AGI triggers, or compute credits. The conversation turns to what this says about negotiating power and the optics of strength versus necessity.
If these were public stocks: long Anthropic, short OpenAI? Employee liquidity decisions
They explore a hypothetical public-markets trade: long Anthropic and short OpenAI to isolate relative execution risk. They also discuss what OpenAI employees should do when tender offers and liquidity windows open at lofty valuations.
OpenAI management shake-up: exits, leaves, and the risks of ‘rebooting’ go-to-market
A wave of senior leadership changes at OpenAI is interpreted as a response to competitive pressure. They focus on the execution risk of bringing in a high-profile operator and handing them a huge scope during turmoil.
OpenAI buys TBPN: bull case vs ‘vanity project’ and why timing matters
They argue both sides of OpenAI’s TBPN media acquisition, then largely converge on skepticism. A key nuance: the deal process likely started months earlier, and management turnover can kill deals that made sense at signing time.
Which would you buy at swapped prices: OpenAI vs Anthropic? Moats, drama, and leadership fit
They run a valuation inversion thought experiment and land on differing views: one sees OpenAI as fixable with focus, another avoids the drama and prefers deeply technical leadership. The discussion highlights consumer distribution, product priorities (e.g., coding), and governance as decisive variables.
SpaceX confidential IPO filing: $2T target, Elon premium, and market mechanics
SpaceX’s rumored IPO and valuation become a lens on today’s extreme power-law outcomes. They discuss how the ‘Elon premium,’ retail demand, float dynamics, and underwriting negotiations could ‘will’ a price into existence—at least initially.
Sequoia’s Doug Leone returns: continuity, competition, and ‘gravitas’ in a transition
Doug Leone’s return to investing (not leadership) is interpreted as a stabilizing move amid firm evolution. They debate whether it’s about LP comfort, founder competitiveness, or simply adding a proven closer back into deal dynamics.
YC kicks out Delve: fraud, community trust, and why ‘breaking the code’ is unforgivable
Delve’s ejection from YC is discussed as enforcement of community norms rather than a surprise occurrence. They distinguish between ‘misusing AI’ and allegedly stealing/forking a batchmate’s work—crossing a line that undermines the YC network itself.
OpenRouter’s rise: the ‘Stripe/Twilio for LLMs’—and the small-take-rate scaling dilemma
They explain OpenRouter as a routing layer across dozens of models that charges a small fee on large inference spend. The product-love is high, but they wrestle with whether a low take rate can credibly scale to $1B+ revenue without expanding into multiple products.
Supabase at $10B: agents creating more databases than humans and the new default stack
Supabase is framed as a pre-AI company that perfectly attached to the agent/vibe-coding boom. They argue databases are foundational, creation is exploding, and Supabase’s ease-of-deploy makes it a natural monetization point—if it can defend against future insourcing/commoditization.
Mercor hack and the coming AI cyber era: why ‘no tolerance’ is the new vendor standard
They unpack the Mercor breach as a preview of AI-amplified cyber risk where automated adversaries can scale attacks cheaply. The key concern is enterprise customers’ low tolerance for vendor security failures in a world where alternatives are plentiful and switching is easier.
The ‘$1.8B two-person company’ lesson: AI-powered marketing at scale (and the dark-arts pipeline)
They treat the headline as both overhyped and revealing: the tactics may live on the edge of regulation, but the operational leverage and hyper-personalized marketing playbook foreshadows mainstream marketing in 12–24 months. The broader claim is that marketing is becoming more powerful, not less, and AI is accelerating the arms race.
