What separates the AI startups gaining real traction from those burning out?

At the AI Rabbit Hole Conference, Benchmark Capital general partner Sarah Tavel joins Reuters tech correspondent Anna Tong to unpack how early-stage investing is evolving in the face of foundation models, rapid churn, and unprecedented competition.

They discuss:
• The shift from software spend to human capital spend
• Why existential risk is higher than ever for AI startups
• What makes a moat in a world with 45-second switching costs
• Which founders win in this environment (and why)
• The timeless “10x better & cheaper” playbook
• The rise of agent-native companies that sell output, not software

Timecodes:

00:02 – Intro and Sarah’s background: Bessemer, Pinterest, Greylock, Benchmark
01:33 – Benchmark’s thesis: escape competition, win by a mile
02:27 – How this AI boom differs from the 2010s SaaS era
04:18 – Software spend vs human capital spend: a bigger market
05:07 – Speed, intensity, and existential risk in the AI era
06:01 – Why AI startups face brutal logo churn and low switching costs
07:45 – Threats from above: will foundation model companies move upstack?
09:26 – Founders matter more than ever: adaptability is everything
10:17 – What to look for: paranoia, ambition, and enduring moats
12:10 – Playbook 1: 10x better and cheaper (DeepL, HeyGen examples)
16:09 – Playbook 2: Sell output, not software (Sierra, 11X)
17:53 – Skeuomorphism in SaaS: moving beyond co-pilot UX
18:44 – Final question: Are we in a bubble? Sarah’s honest take