They wanted investing practice for one summer and stumbled into Y Combinator

On the night of March 11, Paul Graham, a programmer who had sold his startup to Yahoo, and Jessica Livingston were walking home from dinner when they stopped at the corner of Garden and Walker streets in Cambridge.

Livingston, who led marketing for a Boston investment bank, had been interviewing for a marketing role with a Boston venture capital firm, which took weeks to decide. While she waited, Graham started telling her what needed fixing about venture capital.

That sidewalk decision became Y Combinator, the startup school that would fund thousands of young companies in twice yearly batches. The idea at first was not to reinvent investing. It was to get practice.

Graham's prescription, as he tells it, was that firms should "make a larger number of smaller investments instead of a handful of giant ones," back younger and more technical founders instead of business school graduates, and let founders stay as chief executive. Rather than wait for the venture firm to act, Graham and Livingston, who later married, decided to start their own investment firm and test those ideas.

Ignorance shaped what happened next. They did not know how professional investors worked, so they looked for a low stakes way to learn. Undergraduates took summer jobs at tech companies each year, so Graham and Livingston proposed a summer program in which students would start startups instead.

They called it the Summer Founders Program and posted an invitation online. They would use a building Graham owned as headquarters, eat dinner together every Tuesday, and bring in startup experts to speak. The logic, as Graham put it, was that they would be, in a sense, fake investors working with founders who would be, in a similar sense, fake founders. The students would at least have a more varied summer than working at Microsoft.

The response made it real. Graham's essays drew 225 applications, many from people already out of school or about to graduate that spring. They invited about 20 groups to meet for interviews and chose eight to fund. That first group included the founders of reddit, Justin Kan and Emmett Shear, who later founded the video site Twitch, writer and activist Aaron Swartz, and Sam Altman, who later led Y Combinator.

The terms were $6,000 per founder, or $12,000 for a typical two person team, for 6 percent of the company. That was better than the deal Graham had taken for his own startup, $10,000 for 10 percent from an early backer. That first summer was really hot, and Livingston supplied free air conditioners to the teams.

What surprised Graham was how much the batch itself mattered. Funding startups together, Graham wrote, "solved one of the biggest problems faced by founders: the isolation," because each team suddenly had colleagues who understood its problems. As later batches arrived, alumni helped the newcomers, and startups began buying from one another, trade Graham and the founders jokingly called the YC GDP.