Scale with small, sharp teams
Leaders are pushing a model of extreme productivity rather than headcount growth: Zerodha's CTO described scaling millions of users with a 35‑engineer team by sticking to first principles, simplicity and focusing on real problems, while other threads recommend pre-investing about 40% of engineering capacity into infrastructure to avoid later scaling debt. Netflix engineers are already rebuilding large workflow systems (mixing rules, ML and human review) to enable agentic workflows, underscoring that productivity often wins over more hires. (x.com) (x.com) (x.com)
A lot of software leaders spent the last decade treating hiring like scaling, and now some of them are arguing the opposite: a company can add millions of users faster by sharpening systems than by adding managers and engineers. Zerodha’s chief technology officer, Kailash Nadh, said the brokerage kept its tech team at about 35 people even as its customer base grew past 20 million. (expresscomputer.in) (techstory.in) That sounds backwards if you think software breaks the way a factory breaks. More orders usually mean more workers, but software behaves more like a road system: if the lanes, signals, and exits are designed well, far more cars can move without hiring more traffic cops. (expresscomputer.in) Nadh’s version of that idea is old-fashioned engineering discipline. He said Zerodha still makes product and engineering decisions from “first principles,” meaning the team starts with the actual problem in front of users instead of copying whatever architecture or hiring pattern is fashionable. (expresscomputer.in) The company’s growth curve makes the claim harder to dismiss. Nadh said Zerodha took until 2020 to reach 2 million users, then jumped from 2 million to 8 million users in seven months during the lockdown period, while the same roughly 30-person tech team handled the surge on top of foundations built earlier. (expresscomputer.in) The trick was not building for a fantasy future on day one. Nadh said Zerodha did not build for 50 million users upfront, but built systems in a modular way so they could adjust if scale arrived, which is a different bet from hiring a large team to prepare for every possible outcome. (expresscomputer.in) That is where the infrastructure argument comes in. Investor and writer Ivan Landabaso has been pushing a rule that companies should spend roughly 40 percent of engineering time on infrastructure early, because the bill for weak foundations usually arrives later as outages, slow releases, and teams that need extra people just to keep old systems standing. (substack.com) (ivanlandabaso.com) In plain English, infrastructure is the plumbing behind the app. It is the deployment system, the testing setup, the data pipelines, the permissions, the monitoring, and the internal tools that decide whether 10 engineers can move like 30 or whether 30 engineers spend half the week waiting on each other. (expresscomputer.in) (substack.com) Netflix is showing what that looks like at a larger and messier scale. Engineer Pratyusha Singaraju described a rebuild of content workflow systems so rule-based automation, machine learning models, and human review all run inside one framework, with failure isolation and feedback loops designed in from the start. (gitnation.com) That matters because agent-style software does not arrive as one magical model. Netflix’s team said every title moving through its system triggers decisions made by rules, machine learning, and people, and the new architecture is meant to make future artificial intelligence agents feel like a natural extension of the workflow instead of a fragile add-on. (gitnation.com) So the shift here is not “do more with less” as a slogan. It is a more specific bet that a small team with durable systems, clear standards, and fewer layers can outrun a larger team that keeps paying interest on bad infrastructure and coordination overhead. (expresscomputer.in) (techstory.in) (gitnation.com)
