I noticed two completely different ways of programming: 1. Write some small code, run, see the result, adjust, repeat for hours. 2. Sit for hours writing code, modelling with types, writing tests, then spend minutes fixing what you’ve got wrong. Turns out, this deeply influences the technology, programming languages and frameworks you prefer.
176
110
35
2,465
Replying to @ChShersh
The industry is in dire need of some proper prescriptions. Maybe some day we can come together and view "write a regression together before fixing a bug" the same way doctors view "wash your hands before surgery".

Jun 14, 2025 · 10:21 AM UTC

5
4
1
151
Replying to @kerckhove_ts
This always a good idea. Before fixing a bug, make sure you can actually reproduce the bug and you understand why it’s happening. Also, make sure to write the test, to verify that your fix actually fixes the problem.
11
4
1
175
Yeah this is a super cold take. Anyone who disagree should lose their license to practice. Quick note about your phrasing: that test does not verify what you said it would. It _tests_ that.
3
1
25
more replies
Replying to @kerckhove_ts @ChShersh
That is definitely one of top 5 satisfying things in programming for me. Finding the cause of a bug, writing a test with the scenario which fails, fixing it, seeing the test pass, and knowing that the same bug will never happen again without my test catching it.
1
1
Replying to @kerckhove_ts @ChShersh
the thing is that a high % of code's value is experimenting on live users you don't experiment surgery on live users, normally don't bring FAANG type barriers, standards and cults outside of it, it's not what makes indie coders prosper
Replying to @kerckhove_ts @ChShersh
No, not the same thing, that comparison is nonsense. You can only do it for simple stuff. For rest, prescriptions are not possible; that's also why doing proper thing is hard in this job. Even unit tests are not "one fits all" solution, I know you disagree, but you are wrong.
Replying to @kerckhove_ts @ChShersh
I think sometimes there's reasonable consensus on what we 'should' do, but we're much less consistent about doing it because it's time-consuming and effortful. I see that changing with AI-powered coding, though, where the AI can do much of this sort of legwork.
Load more