HTMX

The Productivity Paradox of Growth Silicon Valley treats hiring like a victory lap. In the standard venture capital playbook, raising a seed round mandates aggressive recruitment to "fill out the edges." Vik Paruchuri, CEO of Datalab, argues this approach is fundamentally flawed. Drawing from his experience at DataQuest, Paruchuri observed a startling phenomenon: after reducing his staff from 30 to seven, productivity and team happiness actually increased. This counterintuitive result suggests that most organizational growth doesn't scale output—it scales friction. Why Big Teams Build Slow Traditional scaling leads to specialization, which creates "tiny boxes." When an engineer's entire scope is a single shopping cart button, they lose sight of the broader mission. Paruchuri identifies four primary productivity killers in larger organizations: specialized silos that can't flex, heavy remote-sync processes, middle management bureaucracy, and the "management tax" senior engineers pay to oversee junior staff. In many cases, cutting a three-person team down to one senior generalist actually accelerates the project because it eliminates the need for constant context-syncing. The Generalist Philosophy Influenced by Jeremy Howard of Answer AI, Paruchuri advocates for keeping teams under 15 people. These individuals must be generalists capable of navigating the entire stack—from talking to customers and reading research papers to cleaning data and writing inference code. This eliminates the "lossy communication" that occurs when different teams handle different stages of a product. At Datalab, a team of just three people successfully trained and shipped Surya, a 90-language OCR model with 99% accuracy, by maintaining end-to-end context. Operationalizing Lean AI Building a high-impact, low-headcount team requires specific architectural and cultural choices. Technically, it means choosing "boring" tech like server-rendered HTML, HTMX, and Alpine.js over complex frontend frameworks. Culturally, it requires hiring for maturity rather than years of experience. The hiring process itself must reflect this: Datalab uses a 10-hour paid project to vet candidates' ability to ship. By paying top-of-market salaries to a few elite generalists rather than average salaries to a large department, startups can leverage AI tools to automate the "edges" of their business while maintaining a blistering pace of innovation. The Future of the 15-Person Giant The goal is to scale productivity through compute and AI rather than humans. As AI tools abstract away routine tasks like data pipeline creation, the value of the generalist who can "do it all" grows. While this model requires the discipline to say "no" to certain revenue streams that require heavy manual labor, it preserves the "golden period" of alignment that most companies lose as they grow.