a16z's First Employee on the History of a16z, Venture Capital & Fixing the Government | Scott Kupor

After LoudCloud and Opsware sold to Hewlett-Packard in 2007, Scott Kupor spent 12–18 months in what he called “indentured servitude” to HP. In summer 2008, Marc Andreessen and Ben Horowitz told him they wanted to institutionalize their angel investing and raise a fund; by September, Lehman Brothers had collapsed. Kupor nevertheless told his wife he would leave his stable HP job and move his family from North Carolina to California. He hesitated for months, and Andreessen and Horowitz eventually pressured him to decide. They joked that if he had agreed sooner, the firm might have been called Andreessen Cooper Horowitz; officially, he became its first employee rather than a co-founder.

There was no job description: Andreessen and Horowitz said they would invest, while Kupor had to figure out everything else required to build a firm. His early work included buying furniture at Costco, stocking food, opening bank accounts, hiring finance staff, and creating operational systems. On his first day, they asked him to call capital from roughly 50 limited partners for an investment. Without an a16z email account, he used personal Gmail and requested wire transfers. One LP replied, “Who the fuck are you?” and warned Andreessen and Horowitz that someone might be impersonating them. The incident prompted Kupor to establish the firm’s domain and proper infrastructure.

Kupor said Michael Ovitz, who had served on the LoudCloud and Opsware boards, influenced a16z’s operating model. At CAA, instead of one agent handling every aspect of an actor’s career, Ovitz surrounded the actor with specialists in radio, international deals, books, and other areas. a16z applied the same idea to venture capital by separating the general partner’s investing role from professional networks for recruiting, sales, marketing, government relations, and business development. Its “reverse pitch” told founders that they would receive not only a partner such as Horowitz but an entire team dedicated to making the company successful.

Only months after closing its first fund, a16z partnered with Silver Lake and Michael Dell on a buyout of Skype, despite having presented itself to LPs as an early-stage venture firm. Skype had been neglected inside eBay but continued gaining users through a powerful network effect. A major obstacle was an intellectual-property dispute involving founders Janus Friis and Niklas Zennström; Silver Lake’s Egon Durban helped structure a collaborative solution. Critics called the deal reckless, but a16z sold Skype to Microsoft 18 months later for four times its purchase price, returning roughly half the first fund to LPs and accelerating the firm’s second fund.

Kupor said traditional venture firms rarely marketed because capital had historically been scarce: entrepreneurs needed firms such as Sequoia, Accel, Greylock, or Benchmark more than those firms needed to persuade entrepreneurs. a16z lacked those firms’ decades-old track records, so Andreessen and Horowitz changed the competition by publicly explaining a new category—venture capital plus a services platform. They received criticism on Sand Hill Road for promoting ideas before they were proven, but Kupor argued that creating a category required telling a narrative, just as an enterprise software company must explain its product rather than assume customers will discover it.

Kupor initially ran operations while Andreessen and Horowitz invested, and he felt frustrated that he was not also investing. He eventually recognized that early-stage technology investing required deep technical expertise he did not possess. For about a decade, the centralized firm grew to roughly 500 people, with operating functions and investors reporting through Kupor. As a16z expanded into crypto, healthcare, bio, American Dynamism, and other domains, the firm created specialized investing groups and aligned operating support around them, leaving legal, finance, and investor relations centralized. Kupor resisted losing scope but later began investing in later-stage healthcare, eventually holding about 15 investments and roughly 10 board seats.

After 16½ years at a16z, Kupor became director of the Office of Personnel Management, which serves as the HR and talent agency for roughly two million federal civilian employees. He had studied health policy since high school and worked with Stanford researchers, including Jay Bhattacharya, but said timing also mattered: his children had left home, Andreessen and Horowitz supported the move, and he had the financial freedom to take the risk. His venture experience led him to view every organization as people aligned around objectives and incentives. He believed government’s problem was less universally poor talent than a system that discourages performance, accountability, and experimentation.

Kupor said government employees are trained to avoid congressional scrutiny, Government Accountability Office reports, and inspector-general investigations, so the safest strategy becomes repeating existing processes. He contrasted that with venture capital, where permissionless risk can produce enormous upside, and proposed “permissioned risk”: experimentation sized to the consequences. Social Security or retirement processing, he argued, should tolerate reasonable operational risk because a one- or two-day delay is not a life-or-death event. Without permission to innovate, he said, agencies become slow, customer service deteriorates, and talented people have little reason to join.

Kupor described a system in which 99.7% of federal employees receive ratings of three or higher on a five-point scale, even though that is mathematically implausible for a two-million-person workforce. Among roughly 7,000 senior executives, about 90% receive the top two ratings, weakening the meaning of bonuses and recognition. OPM is moving toward limiting top ratings to 40% of employees and directing at least 60% of the compensation pool to that group. It is also removing tenure-based promotion barriers. Kupor cited software developer Ed, known publicly as “Big Balls,” who would have been rejected for lacking a college degree and five years’ experience despite exceptional ability.

Kupor said only 7% of federal workers are under 30, compared with 23% of the overall U.S. workforce; 44% of federal workers are over 50, versus about 33% nationally. Because many older employees may retire within five to ten years, OPM is recruiting roughly 1,000 software engineers, AI specialists, and data scientists through TechForce. Participants can serve two-year terms, receive technical assessments and appropriate GS-14-level placement, and work on difficult problems at agencies such as NASA or the Energy Department. OPM has partnered with about 40 private companies, including Alphabet, Meta, and Databricks, to provide training and potential private-sector opportunities afterward.

Kupor described OPM’s retirement operation in an underground limestone mine in Boyers, Pennsylvania. For decades, employees mailed paper retirement forms through HR and payroll departments to the mine, where about 800 staff manually calculated benefits; processing could take three to twelve months. The mine holds approximately 400 million pages, including records dating to the 1950s, and a worker might have to retrieve a physical file whenever a retiree called. OPM has moved new applications online, ended incoming paper on June 30, and installed large-scale scanning equipment. Kupor said electronic applications already cut processing time by more than half, with a goal of issuing a healthy application’s retirement check within 15 days.