Pinned Tweet
AI wrote the code. Great. Who checked it?
I built BORK: 8 AI skills for GitHub issues, draft PRs, reviews, fixes, and E2E/browser checks.
Because “looks good to me” needs evidence. Especially from a robot.
github.com/easterncoder/bork…
27
Spot the PHP bug.
$enabled = (bool) ($settings['enabled'] ?? true);
The settings API sends "false" as a string. What value does $enabled get, and where would you fix it?
48
If an agent keeps fixing the wrong thing, give it one real input, the output you got, and the output you expected. Then ask it to explain the gap before editing.
That example often does more work than another paragraph of instructions.
11
If your WordPress plugin changes how it stores settings, test the upgrade with data from the old version. Run the upgrade twice and check that the second run leaves the data intact.
The clean install is usually the easy part.
21
'May original kami.'
'Saan mapapakinggan?'
'...sa next gig.'
Ready nang baguhin ang sagot? DistroKid signup, 7% off: distrokid.com/vip/seven/1338…
My affiliate link; I may earn a commission.
8
My AI coding setup gives the implementation and review to different agents. The review gets the requirements and the diff. I want a fresh read of what the code does before the author’s explanation shapes it.
What do you still check by hand before merging?
11
An AI-generated refactor can be smaller in lines and larger in risk. Before merging, list the behavior callers depend on. Which of those assumptions does the test suite actually check?
18
A useful bug report gives the next developer a starting point: the smallest input that fails, the expected result, and the result you actually got. What detail do you most often have to ask for?
14
A permission check belongs on the server even when the button is hidden. For a WordPress admin action, test the request as a user who cannot see that button. The UI is not the authorization boundary.
17
Before you add another WordPress performance plugin, check whether the slow page is slow only for logged-in users. If it is, measure the uncached request. On a membership site, that’s the request your paying customers actually wait for.
2
16

