#655 – December 07, 2025
it's not what you think
How good engineers write bad code at big companies
10 minutes by Sean Goedecke
Big tech companies often produce sloppy code despite hiring skilled engineers and having resources to do quality work. Sean says the main reason is that most engineers work on systems they barely know. Engineers typically stay only 1-2 years at companies and change teams frequently, but codebases last much longer.
SST to Intercom: The database for startups, enterprises, and everything in between
sponsored by PlanetScale
Get $50 in credits so you can experience first-hand why companies like Cursor, Intercom, and Cash App choose PlanetScale to power their databases. Blazing fast performance, ultra low-latency, unlimited IOPS, 99.999% uptime - all fully managed and available in AWS or GCP. PlanetScale offers the fastest Postgres and MySQL in the cloud.
Why I ignore the spotlight as a staff engineer
11 minutes by Lalit Maganti
This post contrasts two career paths for senior engineers. Sean Goedecke advocates for chasing high-visibility projects and staying adaptable. Lalit chose the opposite approach, focusing on long-term infrastructure work. By staying with developer tools teams for nearly a decade, Lalit built deep expertise and earned trust from internal customers.
AI skeptic to AI pragmatist
12 minutes by Rockford Lhotka
Rockford describes his shift from AI skeptic to AI pragmatist after using tools like Copilot and Cursor in real software development. He explains what he learned about giving AI the right context, treating it as a collaborator, and switching models when needed. He also share practical tips on prompts, git commits, and MCP servers.
On idempotency keys
9 minutes by Gunnar Morling
Distributed systems can't guarantee messages are delivered exactly once, but they can ensure exactly-once processing. This works by adding unique idempotency keys to messages. UUIDs work as keys but require storing all previous keys. Monotonically increasing sequences are better because consumers only need to store the latest processed number. However, this creates bottlenecks for multi-threaded producers. Gunnar suggests a solution which uses database transaction logs to assign sequential keys asynchronously, combining efficiency with performance.
No code reviews by default
5 minutes by Thomas Paul Mann
Thomas doesn't require code reviews at his company. Engineers push to the main branch and request reviews when they think it's necessary. In this blog post, he outlines how they built an engineering culture that is based on trust and allows them to move incredibly fast.
And the most popular article from the last issue was: