TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Software engineer Neil Kakkar says he merged 11 times as many pull requests per month at AI-native startup Tano as he did at PostHog before its AI adoption. His account points to faster individual output alongside new challenges, including context switching and maintaining shared understanding; the figures are one engineer’s experience, not an industry-wide measure.
Software engineer Neil Kakkar says he merged 11 times as many pull requests per month at AI-native startup Tano as he had at PostHog before its adoption of current AI tools, in an essay published by The Pragmatic Engineer. His account describes a shift in how one engineer works: AI agents helped accelerate shipping, while increasing pressure to manage context, focus and review quality.
Kakkar reports that he moved from four years at PostHog to Tano, where he joined as its first engineer and worked for eight months. At PostHog, he says, he shipped about 11 times fewer pull requests per month than at Tano. He also reports shipping 2.4 times as many per month at PostHog as during his earlier work at Bloomberg, where he was starting his career. The essay gives the comparisons as personal output figures, not a controlled study.
The work at the two startups covered a wide range of engineering tasks. At PostHog, Kakkar says he worked on analytics infrastructure, experimentation, feature flags, surveys and early-access products over four years. At Tano, he describes building an affiliate platform, improving a content-monitoring system and migrating the company’s agent architecture in a period of eight months. He identifies AI as a major contributor to the faster pace, alongside the startup’s engineering culture.
The comparison has important limits. Kakkar calls pull-request counts a metric that can be gamed, though he says he did not try to inflate his numbers. He also says PostHog’s figures are a pre-AI baseline: he notes that the company has since embraced AI and that its output is increasing, but the essay provides no comparable post-adoption figures. His Bloomberg data is described as rough.
Faster Shipping, New Workflow Costs
Kakkar’s account offers a concrete example of how AI coding agents may change an engineer’s day-to-day output, while also showing why more code does not settle whether a team is more effective. Pull requests measure a particular activity; they do not by themselves establish code quality, product impact, reliability or the amount of work required from other team members.
He says some former constraints, including communication overhead and extensive code review, diminished at Tano. In their place, he describes context switching, loss of shared understanding, less attention to design and concern about AI plan limits. For engineering teams, the implication is practical: adopting agents can shift bottlenecks rather than remove them. Teams may need to adapt review, coordination and planning practices to keep rapid output understandable and maintainable.
Products Worth Considering
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
Samsung Odyssey G4 25-inch FHD Gaming Monitor
Samsung 27-inch Odyssey G5 QHD Gaming Monitor (G51F)
As an affiliate, we earn on qualifying purchases.
From Tab Completion to Agents
The essay compares workplaces with different sizes and working cultures. Kakkar describes Bloomberg as a large financial technology company with about 20,000 employees and defined processes. He joined PostHog when it had around 10 people, according to his account, and says it had grown to about 200 by the time he left. Tano had around 10 people when he wrote about it, with Kakkar joining as its first engineer.
He characterizes PostHog through late 2024 as a pre-AI workplace where AI use was mostly tab completion and assistance with repetitive tasks. At Tano, he says, agent tools could take on larger tasks, including building features with less direct human input. The comparison therefore reflects both a change in software tools and differences in company stage, team size and responsibilities. Kakkar also cautions that PostHog has since become highly engaged with AI, so its earlier output is not a measure of its current pace.
“I’m shipping 11x more PRs at Tano than at PostHog, where I shipped 2.4x as many as I did at Bloomberg.”
— Neil Kakkar, software engineer, in The Pragmatic Engineer
Limits of One Engineer’s Data
The figures come from one engineer’s account; the essay does not describe an independent audit or controlled comparison. It does not provide enough information to determine how the pull requests compare in size, complexity, review effort or eventual product outcomes. Company changes, role differences and the different time periods may also affect the comparison.
It is also unclear how Tano’s reported pace affects defect rates, maintenance burden or long-term reliability. Kakkar describes concerns around context and design but does not quantify them. His essay says PostHog’s output is now increasing as the company adopts AI, but does not supply current figures that would allow readers to compare the two workplaces on equal terms.
Further Essays, Broader Evidence
The Pragmatic Engineer said it plans to publish a follow-up highlighting runner-up essays from its writing challenge. Those accounts may add other experiences, but the supplied material does not establish whether they will use comparable metrics or cover the same tools and work settings.
For now, Kakkar’s figures should be read as a snapshot of his own work at two startups at different stages of AI adoption. Broader conclusions will require comparable information across teams, including quality, delivery outcomes, review needs and maintenance costs—not just the number of pull requests merged.
Key Questions
What is the main finding in Kakkar’s essay?
Kakkar says he merged 11 times as many pull requests per month at Tano as at PostHog before PostHog’s adoption of current AI tools. The comparison reflects his own work, not the average experience of software engineers.
How long did he work at each company?
He describes spending four years at PostHog and eight months at Tano at the time of writing. He also worked at Bloomberg earlier in his career, but the supplied material does not give a precise duration for that comparison in the summary of his work.
Does the essay show that AI makes all engineers 11 times more productive?
No. The reported figure compares one engineer’s monthly pull-request counts across two different companies and periods. It does not establish an industry-wide productivity effect or measure quality, reliability or product impact.
What difficulties did Kakkar associate with faster AI-assisted work?
He describes context switching, loss of shared context, less focus on design and concern about AI plan limits. He also says teams need to adjust their culture and processes to benefit from AI tools.
Source: newsletter.pragmaticengineer.com
