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 reports merging 11 times as many pull requests per month at AI-native startup Tano as at PostHog, where he worked before AI agents became part of his workflow. His account links the increase to AI tools and engineering culture, but it is one engineer’s experience, not a controlled measure of industry-wide productivity.
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, describing a change in how quickly he can build and ship software with AI agents. His comparison, published as the winning essay in The Pragmatic Engineer’s five-year anniversary writing challenge, offers a detailed individual account—not a controlled study of software-industry productivity.
Kakkar says he joined Tano as its first engineer and worked there for eight months, building an affiliate platform, improving a content-monitoring system and moving the company’s agent architecture to a system with persistent memory, different tools and improved observability. He compares that period with four years at PostHog, where his work ranged from analytics infrastructure and experimentation to surveys and early-access features.
In the essay, he gives a broader comparison: he says he shipped 2.4 times as many monthly pull requests at PostHog as at Bloomberg, then 11 times as many at Tano as at PostHog. He cautions that pull-request counts can be gamed, while saying he did not set out to manipulate the measure. He also describes his Bloomberg figures as rough and presents the PostHog period as a pre-AI baseline.
Kakkar attributes much of the difference between PostHog and Tano to AI, as well as to the engineering culture that supported its use. At PostHog, he says, AI was largely limited to autocomplete and repetitive tasks such as generating test cases. At Tano, he found coding agents capable of handling larger portions of implementation. The source material does not provide independent verification of the figures or a complete account of how the monthly counts were calculated.
AI Changes the Pace of Shipping
Kakkar’s comparison captures a shift facing engineering teams as AI coding agents move from assistance with small tasks toward work that can span a feature. If more code can be produced quickly, teams may need to rethink how they coordinate work, review changes and maintain a shared understanding of a project. The essay’s central point is that adopting tools alone is not enough: workplace practices and team culture also shape how much value engineers get from them.
The reported jump in pull requests should not be read as proof of an equivalent rise in useful software, quality or business results. Kakkar’s account also describes costs that may accompany the higher pace: context switching, weaker shared context, less focus on design and concern about AI plan limits. Those tradeoffs matter to readers evaluating whether faster output means better engineering or simply more work moving through a pipeline.
Products Worth Considering
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
LG UltraGear 27GX704A-B 27-inch QHD OLED Gaming Monitor
Samsung Odyssey G4 25-inch FHD Gaming Monitor
As an affiliate, we earn on qualifying purchases.
From Autocomplete to Coding Agents
The comparison covers three distinct workplaces at different career stages. Kakkar describes Bloomberg as a large financial technology company with defined processes, PostHog as a fast-moving startup that grew from around 10 employees to about 200 during his tenure, and Tano as a small AI-native startup where he was the founding engineer. These differences make the companies useful examples of contrasting environments, but also complicate a direct productivity comparison.
At Bloomberg and PostHog, Kakkar says, code changes required peer review, and senior engineers spent substantial time reviewing others’ work. He describes PostHog before late 2024 as a pre-AI reference point. He also says PostHog has since embraced AI and is increasing output, so his essay does not present the company’s current practices as unchanged.
“I’m shipping 11x more PRs at Tano than at PostHog”
— Neil Kakkar, software engineer, in The Pragmatic Engineer
What the Pull-Request Count Misses
The account does not establish whether the 11-fold difference reflects AI alone. Kakkar moved between companies with different sizes, processes, products and responsibilities, and the comparison also spans different points in his career. The source provides no independent audit of the counts, consistent baseline across all three employers, or adjustment for working hours, project scope or the complexity of each change.
It is also unclear how much of the faster output reached users, how the work compared in reliability or maintenance cost, and how much time was spent correcting agent-generated code. Kakkar flags context loss, design attention and tool-plan limits as concerns, but the essay does not quantify those effects. The reported change is a personal observation, not a claim that all teams will achieve the same results.
More Essays, Broader Evidence
The Pragmatic Engineer said it plans to publish selected runner-up essays in a follow-up issue. That may add other engineers’ experiences to the discussion, but the source does not announce a study that would test Kakkar’s figures across teams.
For now, the next useful evidence would be measurements that look beyond pull-request volume: delivery speed, defects, rework, maintenance, customer outcomes and the time engineers spend reviewing or directing AI agents. Until those measures are available, Kakkar’s account is best treated as a detailed example of how AI tools and engineering culture can change one engineer’s workflow, with the overall effect on software quality and team productivity still unsettled.
Key Questions
What does Kakkar mean by 11 times more?
He says he merged 11 times as many pull requests per month at Tano as at PostHog. The figure is his own comparison and is not an industry-wide statistic.
Does the essay prove AI caused the increase?
No. Kakkar attributes much of the change to AI, but he also changed employers, roles and working environments. The account does not isolate AI’s effect from those other factors.
Were pull requests the only measure of productivity?
The essay uses pull requests as a prominent measure and discusses the work behind them. It does not establish that the count directly measures software quality, customer impact or overall productivity.
What problems did Kakkar associate with faster work?
He describes context switching, loss of shared context, less attention to design and concerns about AI plan limits as costs that can arise when teams move faster.
Source: newsletter.pragmaticengineer.com
