TL;DR
Get comfort and recovery gear 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 at PostHog, where he had worked before. His account, published as the winning essay in The Pragmatic Engineer’s challenge, describes faster shipping alongside new problems including context switching and weaker shared understanding.
Software engineer Neil Kakkar says he merged 11 times as many pull requests per month at AI-native startup Tano as he did at his previous company, PostHog, in a comparison published by The Pragmatic Engineer. His account points to a sharp change in how quickly one engineer can ship work with AI tools, while also describing fresh costs in attention, coordination and design.
Kakkar’s essay was selected as the winner of an essay challenge run by The Pragmatic Engineer for its fifth anniversary. The publication said it received 125 submissions and awarded a $10,000 grand prize to its chosen article. Kakkar, a London-based engineer with nine years of professional experience, described work at Bloomberg, PostHog and Tano.
His comparison is based on his own work history: he spent four years at PostHog and eight months at Tano, where he joined as a founding engineer. He says he shipped 11 times more pull requests per month at Tano than at PostHog; his PostHog rate was about 2.4 times his rate at Bloomberg. He also described his PostHog work as compressed into eight months at Tano, though the roles and time periods were not identical.
Kakkar said AI tools were largely limited to code completion at PostHog until late 2024, while his later work at Tano involved agents capable of building features with less direct input. He attributed the output difference in part to AI, but also stressed the role of engineering culture. Pull-request counts are a personal productivity measure, not a direct measure of product quality, business results or overall team output.
Faster Shipping, New Engineering Costs
Kakkar’s account offers a concrete example of how AI-assisted development may change an engineer’s day-to-day work: more changes can be prepared and merged in less time. For teams, that can make it possible to build and revise software faster, but it also raises questions about review practices, coordination and whether colleagues retain a shared understanding of a changing codebase.
Those tradeoffs matter to developers and employers deciding how to adopt coding agents. Kakkar says the old bottlenecks of communication overhead and code review diminished at Tano, while context switching, loss of shared context, less focus on design and concern about AI usage limits became more prominent. These are observations from one engineer at a small startup, not evidence that every team will see the same results.
The comparison also cautions against treating volume as a complete measure of productivity. A high pull-request count can reflect rapid iteration, but does not by itself show whether software is more reliable, maintainable or valuable to users. Kakkar acknowledges that PR counts can be gamed and presents his figures as an honest account of his work, rather than an independent audit.
Popular right now
Shark HydroVac Reveal Cordless Wet-Dry Vacuum and Mop, Lilac, WD321
Ninja Luxe Café Mini Espresso and Drip Coffee Machine, ES301
Vitamix Propel Series 510 Professional-Grade Blender, Black
As an affiliate, we earn on qualifying purchases.
From Bloomberg to an AI-Native Startup
Kakkar’s career spans three different workplaces. He began at Bloomberg, a company he describes as having about 20,000 employees, where he encountered defined processes and a high bar for peer review. He later joined PostHog, which he says grew from roughly 10 people when he arrived to around 200. He characterized it as a fast-moving startup that encouraged engineers to take responsibility, respond to customers and ship quickly.
He uses PostHog as a pre-AI baseline, not as a description of the company’s current practices. The source report says PostHog has since adopted AI extensively and that its output is increasing, but gives no comparable figures. At Tano, Kakkar joined as the first engineer; he says the company had about 10 people when he wrote the essay.
Across the two startups, Kakkar says he worked on a broad range of products and infrastructure. At PostHog, that included analytics systems, experimentation, surveys and feature tools. At Tano, he describes building an affiliate platform, improving content monitoring and migrating the company’s agent architecture. The work was not identical, making the comparison illustrative rather than a controlled test of AI’s effect.
“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, in his essay published by The Pragmatic Engineer
Limits of One Engineer’s Comparison
The figures are self-reported and specific to Kakkar’s work. The source does not provide an independent review of his pull-request data, a standardized measure of the size or difficulty of each change, or a controlled comparison separating AI use from differences in roles, companies and project timelines.
It is also unclear how much of the increase came from AI agents compared with Tano’s size, processes, engineering culture or the particular work Kakkar was assigned. The essay’s overview mentions limits on AI plans and the risk of losing shared context, but does not provide usage data or quantify how often those problems occurred. The account does not establish whether the higher merge rate led to better software quality or improved business outcomes.
More Essays and Broader Evidence
The Pragmatic Engineer said it would publish a follow-up highlighting runner-up essays and linking to other submissions that stood out. Those accounts may add perspectives from other engineers, although the source report does not say they will use comparable productivity measures.
For now, Kakkar’s essay is a case study rather than a broad verdict on AI-assisted engineering. Evidence from more engineers and teams, using consistent measures for output, review effort, quality and maintenance, would help clarify whether the gains he reports are typical and how they compare with the new coordination costs.
Key Questions
What is the main development in this report?
The Pragmatic Engineer published Neil Kakkar’s winning essay comparing his software work before and after joining AI-native startup Tano. Kakkar reports a much higher monthly pull-request rate at Tano than at PostHog.
How much higher was Kakkar’s reported pull-request rate?
Kakkar says he merged 11 times as many pull requests per month at Tano as at PostHog. He says his PostHog rate was about 2.4 times his Bloomberg rate. These are personal comparisons, not industry-wide statistics.
Does the comparison prove AI caused the increase?
No. Kakkar attributes much of the difference to AI, but the comparison also spans different companies, roles, teams and projects. It does not isolate AI’s effect through a controlled test.
What problems did Kakkar associate with faster AI-assisted work?
He described more context switching, reduced shared context, less focus on design and concerns about AI usage limits. His account does not quantify how often these issues occurred.
What happens after the winning essay?
The Pragmatic Engineer said it planned a follow-up covering runner-up essays and other notable submissions. No broader findings or publication date were specified in the source material.
Source: newsletter.pragmaticengineer.com
