
Q&A with Accel Founder Arthur Patterson

super{set} companies come together every other week to hear from interesting guest speakers from across tech ecosystem as part of our ongoing Community Call series.
For more on Arthur Patterson, click here.
For more on Tom Chavez, click here.
1. On Starting Accel & The Early Days of Venture Capital
2. On Early Stage Hiring
3. On Time Horizons in Venture Capital
4. On Hiring Executives
5. On Company-Building as an Experiment
6. On the Patterson Cycle
7. On Crypto
8. On Why Tom Wasn’t Fired as CEO of Rapt
9. What are you most proud of? (Audience Question)
10. What is the common thread across big hits and misses? (Audience Question)
11. How do you sort the signal and the noise? (Audience Question)
12. Is it problematic that traditional VC funds are getting so big? (Audience Question)
Tech, startups & the big picture
Subscribe for sharp takes on innovation, markets, and the forces shaping our future.
NEWS, BLOGS & ARTICLES
Let's keep in touch
We're heads down building & growing. Learn what's new and our latest updates.

AI didn't eliminate the hardest part of building a startup—it moved it. Today, the challenge isn't writing code; it's finding the people who can validate your idea before you race to market. Design partner discovery has become the new go-to-market problem, and the founders who solve it first will build companies that matter.

Every company I talk to is building agents, AI-native tools, and therefore integrating LLMs into their customer-facing production systems. The pace of adoption has been staggering. OpenAI’s rapid growth and Anthropic’s imminent IPO are clear validation points. But there’s a pattern I keep seeing: the demo works, the executive buyers are excited, the board is bought in—and then it hits production, and everybody is surprised with the LLM costs, in the form of credits or tokens consumed. Often this is because of lack of control and foresight with LLM usage as adoption and usage grows. Often there are little to no guardrails or notifications built in to inform users on the degree of usage and consumption over time. Also, there are no guardrails for engineering teams building the AI native solutions for downstream end users.
