TL;DR
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
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 widespread use of coding agents. His account links the change to AI tools and workplace culture, while also describing costs including context switching, weaker shared context and less attention to design.
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, in a first-person account published by The Pragmatic Engineer. His comparison offers one engineer’s measured view of how AI tools and workplace practices can change software output, while also outlining new pressures that come with faster work.
Kakkar’s figures compare work at PostHog, where he spent four years, with his first eight months at Tano, where he joined as a founding engineer. He says his output at PostHog was 2.4 times his earlier rate at Bloomberg, where he began his career. The essay describes pull requests merged per month as its central measure, but the source does not give the underlying monthly counts in the supplied material.
The work covered at the two startups was broad. At PostHog, Kakkar says he contributed to product areas including data ingestion, analytics on ClickHouse, experimentation and feature flags, and helped build surveys and early-access features. At Tano, he says he worked on an affiliate platform, improved a content-monitoring system and migrated the company’s agent architecture. He characterizes the Tano work as compressing comparable breadth into eight months rather than 45.
Kakkar attributes much of the difference to AI, alongside differences in engineering culture. He says AI at PostHog, through late 2024, was largely tab completion and assistance with repetitive tasks; at Tano, he encountered coding agents capable of carrying out larger tasks. He cautions that pull-request counts can be gamed, and presents his figures as an honest record rather than a controlled study of AI’s effect.
Higher Output, New Workflow Costs
The account matters because it gives a concrete, though individual, example of the productivity shift software teams are trying to understand as coding agents spread. A reported rise in merged pull requests may interest engineers and employers weighing whether AI changes how much work a small team can deliver. It does not, by itself, show that the software was better, that users received more value, or that AI alone caused the increase.
Kakkar also describes tradeoffs behind the higher pace: more context switching, loss of shared context, less focus on design, difficulty pacing work and concern about AI plan limits. He says older constraints such as communication overhead and extensive code review have receded in his current environment, but that moving quickly can mean breaking things. The implication is that adopting agents is not just a tooling decision; teams may need to rethink review, coordination and how they judge engineering output.
Products Worth Considering
Apple Watch SE (2nd Gen) GPS 40mm with Starlight Aluminum Case and Sport Band (S/M)
Samsung 27-inch Odyssey G5 QHD Gaming Monitor (G51F)
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Black
As an affiliate, we earn on qualifying purchases.
From Tab Completion to Agents
Kakkar’s career provides the comparison behind the essay. He worked at Bloomberg, a large fintech company, then PostHog, which he says grew from around 10 employees when he joined to about 200, and then Tano, where he was the first engineer and the team later reached around 10 people. These workplaces differed in scale and process as well as in access to AI, so the comparison spans more than a change in tools.
The Pragmatic Engineer published the essay as the winner of a challenge marking the newsletter’s fifth anniversary. The publication says it received 125 submissions and awarded the winning article a $10,000 grand prize. Kakkar describes PostHog as a pre-AI baseline for the comparison, while acknowledging that the company has since adopted AI tools and that its output is increasing, according to his account.
“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
What The Comparison Cannot Show
The essay is a single engineer’s retrospective, not a controlled study. The supplied source does not include the exact monthly PR counts, consistent measurement rules, or independent checks of the work represented by each pull request. It also does not quantify changes in code quality, defects, customer outcomes or the amount of work completed by colleagues.
Because Kakkar moved between companies with different sizes, products, responsibilities and processes, the comparison cannot isolate the effect of AI from those other factors. He says AI was a major contributor, but the size of its independent effect remains unclear. The account also does not specify which tasks agents completed without human intervention or how much review and correction their output required.
More Essays, Broader Evidence
The Pragmatic Engineer said it planned to publish selected runner-up essays in a follow-up issue and link to other submissions that stood out. Those accounts may offer additional perspectives, though they would not on their own establish a general productivity effect.
For now, Kakkar’s comparison is best read as a detailed report from one engineer across three workplaces. Broader evidence would need to compare output alongside quality, delivery time, maintenance costs and customer impact, while accounting for differences in team structure and project scope.
Key Questions
What did Neil Kakkar report?
He said he merged 11 times as many pull requests per month at Tano as at PostHog. He also reported that his PostHog rate was 2.4 times his earlier rate at Bloomberg.
Does the comparison prove AI caused the increase?
No. Kakkar attributes much of the difference to AI and engineering culture, but this is a personal comparison across companies with different teams, processes and responsibilities. It does not isolate AI’s effect.
What drawbacks did he describe?
He points to context switching, lost shared context, reduced design focus, difficulty pacing work and concerns about AI plan limits. He also describes the risk of moving faster and breaking things.
Is a pull-request count a measure of software quality?
Not by itself. Kakkar says pull-request counts can be gamed. The essay does not provide comparable measures of code quality, defects, customer impact or maintenance work.
Source: newsletter.pragmaticengineer.com
