Steve | Systems & AI
Region: US
Wednesday 30 September 2026 19:25:55 GMT
4179
116
35
3
Music
Download
Comments
labtomarket :
test coverage has nothing to do with the code review and was always automated
2026-10-01 17:27:05
0
SedimentLayerAnalysis :
We'd all check eachothers code like we all thought the other was a Russian spy
2026-10-01 08:07:14
0
burner_acct_03 :
what about static analysis? do we care about that anymore?
2026-09-30 23:25:26
1
kpoyaka :
have you guys consider the absence of those Ai tools in the future?
2026-10-01 07:00:05
1
chiefsFanSixtySeven :
No way a person or team can manually review the amount of code generated by AI. To be fair back in the day most of the manual code reviews weren’t that detailed if you trusted the developer submitting the pr. For junior/new devs you’d definitely give the pr’s more attention
2026-10-01 02:45:49
1
Alex :
I want to buy your systems, can I buy the first option, and upgrade after paying the remaining amount?
2026-10-01 03:44:51
1
PhoenixStalfos :
AI wrote my tests too. lmao
2026-10-01 02:17:28
4
jayemathisen :
I use a completely different ai system to make sure code matches spec then run the tests with a third. it's worked well. make them all very adversarial and then I review the high priority findings.
2026-10-01 04:59:07
1
user2061828755237 :
But then you have to read the tests. If you don’t read the tests, then the code is working perfectly doing the wrong thing
2026-09-30 19:38:01
3
Aur Naur A Divergent :
Eh. Idk. I see it as a choose 2 out of 3 thing. You can have ai write the tests. You can have ai write the code. You can have ai review the code. We are already saving so much time with 2 out of 3. Reviewing is probably what makes the most sense to invest human effort in, because that’s where we invest in scalability and readability and maintainability. If we aren’t at least engaged in review, there’s no real benefit in having someone technical behind the wheel at all. You might as well get rid of engineers if you hand it all over to ai. I can’t really imagine choosing to have the test writing be the human part.
2026-09-30 23:13:09
1
ResistForHumanity0 :
I write the festure...then the tests. if updating a feature the ai fixes bugs tests caught or updates and adds if test coverage was outdated.
2026-09-30 20:53:05
1
abdesselam :
Tests only check what someone thought to test. If the AI wrote the code AND the tests, green just means they agree with each other. And tests inside one service can't catch a contract breaking between services. Service A changes a field, B's mocks still pass, prod goes down. You don't read every line, but you still review the spec, the contracts, and the risky stuff (auth, payments, data). Tests are a safety net, not a reviewer.
2026-09-30 23:14:57
1
tiktoker1001 :
This matters for porting to new language and having ai rewrite the whole code
2026-09-30 19:41:24
1
tylerdurden :
I begrudgingly agree with this. The purest in me is resisting.
2026-09-30 21:00:09
2
Omulangira Kalema :
I have used Claude to write code, logic, process, and user tests. Truth is developers will soon be out of every job. I use Claude in the browser to do every single thing I need for shipping, from Git, aws (I use Oracle Cloud), cloudflare... everything.
2026-09-30 20:41:30
1
tiktoker1001 :
Testing is also decoupling, fuzzing corpus is also valid
2026-09-30 19:42:07
1
tiktoker1001 :
Don’t call it test coverage but spec coverage
2026-09-30 19:43:11
1
mitchnewhouse7 :
Software projects only need 2 humans. 1) Product Owner to gather and input requirements to AI agents to generate code. 2) QA to input testing criteria for AI agents to verify.
2026-10-01 16:13:00
0
To see more videos from user @sd141414, please go to the Tikwm
homepage.
![vou colocar - remix [overlapped] MC menor K, DJ ery | #fyp #edit #audio #audiosforedits #funkedit](/video/cover/7637601414570708244.webp)
