The 'It Works On My Machine' Pull Request
Description
A two-panel meme using the 'Drifting Car Exit' format. In the top panel, a green highway sign points straight ahead with the text 'SUBMIT A PULL REQUEST'. An exit ramp to the right lists a series of software development best practices: 'WRITE UNIT TESTS', 'TEST FEATURE AGAINST PROD DATA', 'ACCOUNT FOR EDGE CASES', 'REFACTOR FOR MAINTAINABILITY', 'REFACTOR FOR READABILITY', and 'GIT SQUASH ALL OF YOUR MEANINGLESS COMMIT MESSAGES'. The bottom panel shows a blue car recklessly drifting, tires smoking, to take the exit ramp, which is labeled 'MY COWORKER WHO SOMEHOW GOT THEIR CODE TO RUN ONCE ON DEV UNDER IDEAL CONDITIONS'. A watermark for 't.me/dev_meme' is in the bottom-left corner. The meme satirizes developers who, after a single successful test on their local or a development environment, rush to merge their code while completely bypassing essential quality assurance steps. This behavior is a classic example of the 'it works on my machine' fallacy, creating technical debt and instability that more experienced engineers often have to resolve
Comments
7Comment deleted
The fastest way to get your code into production is to make sure it only has to work correctly once, and only on your laptop. The SRE team can handle the rest
Nothing says Friday afternoon like a 4,000-line PR from the “ran once on dev” crew - no tests, seven commits titled “fix”, and my pager whispering “see you at 3 AM.”
"Works on my machine" evolved from a junior excuse to a senior's containerization strategy, but somehow Jenkins still finds a way to make it fail
We've all worked with that developer who treats 'it compiled once on localhost' as sufficient QA, then submits a PR that's essentially a game of production roulette. They've discovered the ancient art of transforming technical debt into someone else's problem - usually yours during the 3 AM incident call when their untested edge case decides to surface in prod. The real tragedy? Their code somehow passes CI because the test suite is as comprehensive as their commit message 'fixed stuff'
Ran once on dev with a warm cache and fake fixtures; submit PR, git-squash the evidence, and let CI write the postmortem
That PR which 'ran once on dev': the architectural equivalent of a monolith that scales until the first Black Friday spike
Submitting a PR after a single happy‑path run is event sourcing for bugs - append‑only pain with eventual consistency to a 3am rollback