How software engineering is changing: a tale of two eras

Open original

Neil Kakkar, a London engineer, compares his output at PostHog (pre-AI) with his current AI-native startup Tano, where he ships about ten times as many pull requests per month. He argues the real gains come from removing old bottlenecks like review queues and cross-team communication, which AI makes less necessary, while creating new problems around context switching and focus.

  • PR volume went from roughly 8-20 merged per month at PostHog to 200+ per month at Tano, with much of the gain attributed to agents that build full features in 1-2 hours.
  • Mandatory peer review and multi-person feature work were bottlenecks; with AI, one engineer can own a feature end to end from design to backend, which cuts communication overhead.
  • The new constraints are context switching from running 3-5 agent threads at once, losing shared team context, and pacing yourself when output is so fast.

Neil Kakkar, a London-based engineer with nine years of experience, won a $10,000 essay challenge run by Gergely Orosz's Pragmatic Engineer newsletter with a detailed look at how his output changed when he moved from pre-AI workplaces to an AI-native startup. At Bloomberg, where he spent three years at a company of roughly 20,000 people with heavily defined processes, and then PostHog, a fast-moving open source analytics company that grew from about 10 people to 200 during his four years there, he merged between eight and 20 pull requests a month. At Tano, a roughly 10-person startup where he was the first engineer, he ships over 200 PRs a month, about eleven times his PostHog rate. He is careful to say the numbers are honest, not gamed, and he backs them with detailed tracking of his own work. Most of the gain comes from AI, specifically agents like Claude Code Desktop that can build a full feature end to end. Kakkar typically runs three to five tasks at once, lets an agent run for one to two hours until it has a PR with comments addressed, then reviews the output through an HTML artifact the agent generates for each change. He estimates that a pre-AI PostHog with mandatory reviews, transplanted with him and AI tools from the future, would only give him a one to two times gain, since the bottleneck shifts to review and coordination. Much of the rest of the multiplier, in his account, comes from things that disappeared: heavy communication overhead, mandatory peer review of everything, and interruptions from bugs and small requests, which can now be handed to a parallel agent thread instead of breaking focus. The trade-off is that the old constraints were doing real work, and their removal creates new problems. Kakkar lists context switching and thrashing across many parallel threads, losing shared context with teammates, the difficulty of pacing yourself when output is so cheap, less time spent on design, and anxiety about hitting AI plan limits. He also argues that teams need everyone on board with AI before the benefits materialize, that encouraging experimentation matters a lot, and that many existing tools and processes break under the new pace. The essay is a practitioner's account rather than a forecast, and its strongest claim is that the speed gain is real but only as good as the culture and workflow around it.

Ask about this article