The bar for raising a 95th percentile seed round has gotten 3x higher

3 trends in startups/tech/venture

📈 The bar for raising a 95th percentile seed round has gotten 3x higher

• Carta analyzed the ~6,500 seed rounds raised by US startups between 2022 and 2026 and found that valuations have skyrocketed over the last 4 years:

  

• Since investors only get their money back if the startups they invest in “exit” (IPO or sell) at a higher share price than what an investor invested at, more investors are understandably starting to question whether these seed valuation levels are healthy and sustainable.

🤖 Fewer people using more AI is becoming more expensive than more people using less AI

• People used to think that AI could cut headcount and therefore costs. However, more companies are now coming to the opposite realization: they’re spending more money despite employing fewer people.

• How? Because AI isn’t cheap to run—and fewer people using more AI is actually more expensive than more people using less AI.

• Some people have been drawing the analogy that using old-school software is like going to see the movies, while using AI is like going to see live theatre. The latter may be a better experience, but costs a lot more.

💰 Market timing is the biggest driver of VC fund performance

• Per the latest data from StepStone Group’s analysis of over 1,000 US VC funds over 23 vintage years, 80% of venture returns are driven by just 22–30% of vintages*. (*Vintage = the year in which a fund began, so a fund that made its first investment in 2026 is a 2026 vintage year fund.) In fact, over the 23 vintage years analyzed, 80% of returns came from just 5-7 vintage years and 95% of returns come from 6-10 vintage years.

• The high-performing fund vintages? The ones that deployed capital after market corrections—not unlike public market investors who could “buy low” and then “sell high.”

• This resurfaces the age-old debate: are the best investors truly talented … or were they just lucky (and, specifically, lucky in terms of when they started writing checks)? It’s nearly impossible to “time the market,” of course, so the only thing General Partners (AKA venture capitalists) and Limited Partners (AKA investors in venture funds) can do is invest consistently, even when market conditions are bad.

2 theses on what’s next

🤖 There will be a new bullet point in every job description

• As people continue to outsource tasks and workflows to AI agents, there will be a new addendum at the end of everyone’s job titles: “agent manager.”

• If you’re in sales, you will now also be a “sales agent manager.” If you’re a coder, you will now also be a “coding agent manager.” Even if you’re in a C-suite leadership role, you will now also be expected to manage AI agents who are quietly doing a part of your job while you’re sleeping.

• I just had the fortune of hosting KC, OpenAI’s Chief People Officer, for a Jeffersonian dinner, where we discussed this. More notes are coming soon!

🤔 The best startups will build software that’s accountable

• As NVIDIA CEO Jensen Huang previously said, the financial markets "got it wrong" about AI killing SaaS: So far, agents aren’t replacing software; they’re using it — and then letting humans step in as the “neck to choke” if things go wrong.

• The next era of companies won’t just do work (or churn out AI slop); they’ll be accountable for the outcome.

• This is one of many topics I’ve taught in the Berkeley Learn2Launch program (more details here) and SCET Entrepreneurship program (more details here). I’ve also discussed this at the Venture Revolution Conference (more details here).

1 tool I love

🎓 AI-powered financial modeling co-pilot

• Filot takes financial modeling work that used to take humans hundreds of hours and does it in seconds. It even learns each firm’s unique style guide. The co-founders are former students of mine from my UC Berkeley entrepreneurship course (more details here).

• Filot is now serving some of the largest funds in the world, including Tiger Global, Ares and Two Sigma among others. They also just announced major partnerships with data platforms including TresVista (more details here).

More of my work 👇🏼

🎤 Decode videos | For top Berkeley and Stanford founders

• DECODE is the largest founder community co-hosted across UC Berkeley and Stanford. The DECODE annual conference focuses on helping founders in their earliest stages of starting a startup.

• I’ve had the honor of serving on the DECODE board since 2016. You can find videos from the latest conference I helped host here 👇🏼

🤖 Founders' prompt library

• My team and I have crowdsourced the largest AI prompt library optimized for founders – especially fractional founders — to help them build, sell and operate 10x faster and better.

👥 Pop up board

• I regularly host pop-up boards, a unique 45-minute session where founders ask their toughest strategic questions and get tailored advice from top builders and operators from Google, Microsoft, Meta and more.