TL;DR
Software engineer Neil Kakkar reports merging 11 times as many pull requests per month at AI-native startup Tano as he did at PostHog before widespread use of coding agents. His account points to faster output alongside new strains, including context switching and less time for design; the figures describe one engineer’s experience, not a controlled industry-wide comparison.
Software engineer Neil Kakkar says he merged about 11 times as many pull requests per month at AI-native startup Tano as he did at PostHog before the spread of coding agents, according to an essay published by The Pragmatic Engineer. His account offers an individual comparison of how AI-supported work can change engineering output, while also describing costs such as more context switching and reduced attention to design.
Get wellness gear delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Kakkar’s comparison draws on his work at PostHog, where he spent four years, and Tano, where he joined as a founding engineer. He says the Tano work he describes was completed in eight months, compared with 45 months for the work period he uses at PostHog. The source summarizes his monthly pull-request rate as 11 times higher at Tano. It also says his PostHog rate was 2.4 times his earlier rate at Bloomberg, where he began his career.
At PostHog, Kakkar worked across product engineering, including analytics infrastructure, experimentation, feature flags and surveys. At Tano, he says he built an affiliate platform, improved a content-monitoring system and migrated the company’s agent architecture to a system with persistent memory, new tools and more observability. He describes using AI agents that can take on larger tasks than the tab-completion tools available during his earlier PostHog work.
The publication says it received 125 submissions to an essay challenge marking its fifth anniversary and selected Kakkar’s piece as the winner. The article presents his figures as a personal account, not a controlled study. Kakkar also cautions that pull-request counts can be gamed, while saying he did not set out to inflate his own numbers.
Higher Output, Different Bottlenecks
Kakkar’s account matters because it shifts the discussion of coding agents from what they can generate to how their use affects the whole engineering workflow. A higher pull-request count may reflect faster implementation, but it does not by itself establish better software, greater customer value or improved team productivity. The comparison is a signal from one engineer’s experience, not a measure of what other teams should expect.
He says some older constraints, including extensive review and communication overhead, have eased in his current environment. But speed brings other pressures: frequent context switching, difficulty maintaining shared understanding, less focus on design and concern about limits on AI plans. Teams adopting agents may need to reconsider review practices, coordination and how they judge output, rather than treating code volume as the main measure.
Products Worth Considering
Fitbit Sense 2 Smartwatch – Shadow Grey/Graphite
Bowflex SelectTech 840 Adjustable Kettlebell
Blueair Signature SP1i Smart Air Purifier
As an affiliate, we earn on qualifying purchases.
Three Workplaces, Two Eras
Kakkar describes a career across three different engineering environments. He began at Bloomberg, a large financial company, then joined PostHog, which grew from roughly 10 employees when he arrived to about 200. He later became Tano’s first engineer; the startup had around 10 people at the time of the report. Differences in company size and working culture are part of the comparison, alongside changes in AI tools.
He characterizes PostHog through late 2024 as a pre-AI baseline, when AI use was largely limited to tab completion and routine assistance such as generating repetitive tests. At Tano, he says, agent-supporting tools were already available when he joined. The source also notes that PostHog has since adopted AI extensively, so the comparison does not describe the company’s current output against Tano’s current output.
The article’s “two eras” framing is therefore Kakkar’s interpretation of his own experience: a period of conventional engineering workflows and a later period in which agents can take on broader tasks. The time periods, teams and roles differ, making the figures informative as a case study but unsuitable as a direct industry benchmark.
“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 and author of the essay
What the Pull-Request Count Shows
The report does not provide a standardized measure of code quality, defect rates, customer outcomes or total engineering productivity across the three jobs. Nor does it isolate AI’s effect from differences in company size, responsibilities, experience, project scope or team practices. Kakkar acknowledges that pull-request counts can be gamed; the article says he considers his own count an honest record, but it does not independently audit the data.
It is also unclear how much of the reported increase would apply to other engineers or organizations, or how the result would compare if both companies used similar tools and workflows. Kakkar says PostHog’s output has increased as it has adopted AI, but the source provides no comparable current figures. The essay therefore documents one engineer’s experience rather than proving a general productivity multiplier.
More Essays and Wider Comparisons
The Pragmatic Engineer says it plans to publish a follow-up highlighting runner-up essays and linking to other entries that stood out. Those accounts may offer additional perspectives on how roles, practices and workplace expectations are changing, though the source does not say they will use comparable measurements.
For now, Kakkar’s account leaves a broader question for engineering teams: whether faster production of code translates into more useful, reliable software once review, coordination and design work are counted. Answering that requires evidence beyond monthly pull-request totals, including consistent comparisons of quality, delivery time and outcomes for users.
Key Questions
What change does Neil Kakkar report?
He says he merged about 11 times as many pull requests per month at Tano as at PostHog during the periods compared. The figure comes from his personal work records.
Does the report prove AI makes engineers 11 times more productive?
No. It is an individual comparison across different companies, roles and time periods. Pull-request volume does not measure code quality, customer impact or overall team productivity on its own.
What problems does Kakkar associate with AI-supported work?
He describes context switching, weaker shared context, less attention to design and concern about AI plan limits, alongside faster implementation.
Why is PostHog described as a pre-AI baseline?
Kakkar’s comparison uses his PostHog work through late 2024, when he says AI was mostly used for tab completion and routine tasks. The source says PostHog has adopted AI extensively since then, so the figures do not compare both companies’ current workflows.
Source: newsletter.pragmaticengineer.com
