#259 – September 14, 2025
Elliot went from never coding to becoming one of their top contributors
Agentic AI has changed my career
12 minutes by Elliot Graebert
Elliot went from never coding to becoming one of their top contributors using AI tools as a manager. He started with AI-assisted coding for simple bug fixes, then scaled up by running multiple AI agents simultaneously on different problems. By automating workflows with agentic AI tools that can work independently, he now submits up to 10 pull requests per hour while handling bigger tasks like complete code migrations.
Less time digging. More time doing.
sponsored by Unblocked
Unblocked augments your code with the scattered knowledge in GitHub, Slack, Confluence, Jira, and more, so you can find the context you need without having to interrupt a teammate.
Solve customer problems
9 minutes by Mike Fisher
Companies succeed by solving real customer problems, not just studying customer demographics. The Jobs to Be Done Theory suggests focusing on what customers are trying to accomplish rather than who they are. Airbnb started by solving a simple problem: expensive hotels during a conference. Amazon obsesses over customer needs and works backward from there.
Engineering excellence starts on edge
2 minutes by David Heinemeier Hansson
The best engineering teams run unreleased software versions in production to help improve the tools they use. This sounds risky, but bugs are found and fixed by programmers through good testing. Major companies like Shopify and GitHub already run Rails 8.1 beta code in production almost a year before release. This approach creates elite engineering teams that actively shape their tools rather than just consume them.
How do AI-assisted code fixes impact code review time?
3 minutes by Lizzie Matusov
AI code assistants can automatically generate fixes for reviewer comments during code reviews. Meta tested this with thousands of engineers and found an unexpected problem. When reviewers could see the AI-generated fixes, they spent 6.7% more time reviewing code because they felt they needed to verify the AI's work. The solution was simple: show AI fixes only to the code authors, not the reviewers.
Time allocation ≠ capacity allocation
5 minutes by John Cutler
John explains that companies often confuse time allocation with capacity allocation, leading to wasted money and poor decisions. Time allocation simply tracks where team members spend their hours, like dividing 200 hours among five people. Capacity allocation is more complex, focusing on a team's actual ability to deliver outcomes based on factors like codebase quality, team dynamics, and past investments. Most time tracking misses hidden work like waiting, context switching, and firefighting that reduces real productivity.
Refactor vs rewrite and the many paths in between
sponsored by Test Double
It’s often wise to avoid an expensive, time-consuming, big bang rewrite. AI tooling and automated testing can speed up legacy app modernization, making it iterative and less painful. A right-sized subsystem design with seam-based modernization can be useful. There are also infinite paths in between … but sometimes a rewrite does make sense. So how do you decide what’s the right decision for you, your org, your systems, and your team? Find out more about our approach to legacy modernization, including helpful content.
And the most popular article from the last issue was: