Data from Carta shows a typical founder’s stake drops to around 5% after a Series D or later, even lower with co‑founders. That translates to roughly $12‑$25 million on a median $71 million exit, or only $3‑$4 million when you factor in the 15% acquisition rate.
What was a $1M max pre‑seed round in 2014 has ballooned to $10M+ today, simply by rebranding. Founders label huge early financings as ‘pre‑seed’ to keep a later, larger seed round on the books, stretching the term’s original purpose and confusing investors.
Leah Tharin shows that AI‑driven ads and onboarding can push the raw activation rate up, but the share of users who stay past two months plunges. Most dashboards only surface the first number, hiding a hidden churn risk that can erode long‑term growth.
Hardware founders can’t rely on engineering alone; the next six functions, Manufacturing, Deployments, Supply Chain, Finance, Sales, Policy and Marketing, determine whether a product scales. a16z’s deep‑dive shows how to spot talent with cross‑industry instincts and cultural flexibility, and why mis‑hiring in any of these roles stalls growth.
Luke Kanies, founder of Puppet, attended the Local First Conference and tested ATProto for a decentralized review platform. He finds the protocol’s current design blocks data ownership and interoperability needed for apps that could replace Yelp, Goodreads, and similar services. He concludes ATProto isn’t yet ready for a truly public, community‑driven app ecosystem.
Chamath warns that many enterprises see AI token bills double every 45 days while productivity climbs only 5‑10%. He urges CEOs to route cheap models to most tasks, safeguard proprietary data, and consider in‑house AI to avoid waste and shape the $1.4 trillion infrastructure spend. The payoff is tighter ROI and strategic control.
New manager Orin C. Davis advises that before assigning ownership, teams must first agree on four concrete criteria: what counts as finished, what quality looks like, the exact deadline, and which resources are allowable. This simple pact cuts frustration, clarifies expectations, and boosts delivery reliability.
Jason Lemkin notes that AI isn’t a death knell for SaaS, but it will kill vendors that haven’t shipped anything new in five years, especially narrow point‑solution tools. In the AI era, cheap in‑house builds can replace stagnant products, reinforcing the 90/10 rule: buy when you can, but keep innovating or be replaced.
Ramp’s July spend data shows AI support bots now the clearest enterprise AI use case, delivering measurable ROI via ticket‑resolution metrics. At the same time, open‑source Chinese models like DeepSeek and Fireworks AI are creeping onto the vendor list, hinting at the first real commoditization of AI models.
You’re caught up.
That was all 9 stories in the 24 July issue.