The speaker argues that the exponential growth in LLM adoption has created the most fertile ground for startups in over a decade. He predicts the creation of multiple trillion-dollar companies within the next few years, driven by AI's ability to address massive markets.
The speaker bluntly dismisses the corporate narrative of AI as a mere augmentation tool, asserting it will fully replace the vast majority of knowledge work. He predicts that only 1% of current computer-based work will exist in a decade, leading to a massive societal shift.
A new technological paradigm is emerging where AI agents will become the primary interface for business-to-consumer interaction, supplanting websites and mobile apps. This transition will disrupt incumbents who fail to adapt and create opportunities for startups to build new, agent-native companies.
Traditional SaaS investment heuristics, like revenue milestones signifying a category winner, are no longer applicable in the fast-moving AI landscape. Benchmark's diligence now focuses on whether a company's moat strengthens as underlying AI models improve, and the investment thesis targets companies replacing labor rather than just selling software.
The most interesting development in AI is the shift to a third stage of model training focused on reinforcement learning. This involves using deep human experts to create difficult problems and evaluation rubrics, which models then use to train themselves, creating a flywheel for continuous improvement.
Keep pulling the thread on Victor Lazarte.