Keep pulling the thread on Menlo Ventures.
Menlo Ventures and Anthropic launched a $100 million "Anthology Fund" in early 2023 to invest in companies within the AI ecosystem.
The speaker claims that almost no iconic company in venture capital history has originated from a VC's pre-existing investment thesis.
The speaker argues that any enterprise software product that can adopt a product-led growth (PLG) model must do so, or it will be disrupted by a PLG competitor.
The effectiveness of LLM scaling laws is limited by the finite amount of internet data, which means arbitrarily increasing model size no longer yields proportional performance gains.
Khosla Ventures' investment in OpenAI was not based on a specific thesis about LLM chatbots but on the belief that the founders were smart and working on something potentially valuable.
The startup Good Fire is developing mechanistic interpretability for AI models to understand their internal workings, rather than just evaluating their outputs.
Good Fire was founded by former researchers from the interpretability teams at Anthropic, DeepMind, and OpenAI.
The sycophantic behavior exhibited by GPT-4o was a flaw not caught during the evaluation phase, highlighting the limitations of current empirical AI testing methods.
The speaker believes AI will be critical for industries facing labor shortages due to declining birth rates, such as insurance, trucking, and logistics.
Menlo Ventures operates as a low-volume investment firm, with partners aiming to do fewer than five deals per year to ensure they have time to actively support portfolio companies.
Menlo Ventures has hired several tenured operators, including Tim Tully (former CTO of Splunk) and Joff (former CPO of Atlassian), to provide hands-on support to its portfolio companies.
According to a recent Carta report, Series A startups with "AI" in their name or description receive a 30% valuation premium.