How To Go From Researcher to Founder

Francois Chaubard opens by contrasting academic vanity metrics—grades, citations, and conference talks—with the founder’s question: what do people actually want, adopt, and pay for? He says an impressive algorithm is worthless if customers will not use it. He then distinguishes interesting research from important problems: researchers are rewarded for explaining mechanisms, but founders must accept “it just works” when that is enough for users. Product-market fit requires working backward from customer importance rather than pursuing intellectual interest for its own sake.

Chaubard argues that academia’s demand for novelty can mislead founders: a paper is rejected if it is not original, whereas startups can succeed by solving an old problem better. He contrasts Joe Lonsdale’s MIT-style startup—built around a novel patented invention—with a Stanford-style company that starts from one ordinary customer problem and iterates relentlessly. Instacart did not invent online grocery delivery; it executed that problem successfully for 15 years. He similarly cites Lyft/Uber, Anthropic/OpenAI, and Rappi/Instacart, invoking Steve Jobs’s advice to start with customer experience and work backward to technology.

Academic work offers a finish line: a paper submission, an exam, or another moment when the researcher can stop. Chaubard says startups offer no such catharsis; founders need a long-term vision and must work on the same core problem every day, potentially for decades. He criticizes researchers’ tendency to pursue five ideas at once—a useful breadth-first strategy in research—as “smearing the peanut butter too thin” in a company. When founders get bored and switch direction, they also distract the team from its mission.

Chaubard says product-market fit comes from repeatedly talking to customers, not solving a presumed problem in isolation or asking investors for validation. At Focal Systems, his team spent three months collecting retail training data and building models to track where shoppers were in stores; retailers had all called people-tracking a major problem. The resulting system merely showed that shoppers spent time at checkout and the deli counter at lunchtime. When he asked whether stores would pay, they said no because the information created no value; manual video labeling would have produced the same result. He now recommends an AI company begin with a human in the loop before spending millions on compute and training.

Chaubard says scientific exactness should not govern a sales pitch: Geico’s “15 minutes could save you 15% or more” slogan is deliberately imprecise, but founders should use compelling language without lying. He pairs this with Jeff Bezos’s 70% rule: most decisions should be made with roughly 70% of the desired information, because waiting for certainty lets competitors launch first. In the database wars, Michael Stonebraker’s technically superior Quell lost to Larry Ellison’s faster push of SQL; Ellison’s attitude was that Quell might be better, but SQL would win. Speed, Chaubard concludes, can beat perfection.

Chaubard warns academics not to hire an MBA merely to handle sales and business while they continue doing research. A startup CEO owns team management, cash runway, hiring, firing, marketing, sales, and product development; avoiding sales prevents the customer empathy needed to build a wanted product. He also says founders must adopt a more aggressive capitalist psychology—charging money, winning talent, raising capital, pursuing strong margins, and beating competitors—despite academic discomfort. Finally, intelligence alone does not sell: founders should listen to users, whether they are IT staff, interns, procurement officers, or consumers. His practical closing advice is to leave the office, fly to users, test willingness to pay, and read Spin Selling, High Output Management, The Seven Habits of Highly Effective People, and How to Win Friends and Influence People.