There is a lot of controversy about whether agentic coding (LLMs writing code) is a curse or a blessing, rightfully. Let me add an anecdote.
Recently I discovered SitePing, a website annotation tool, and applied it to a prototypical website for reviewers to give direct feedback. It had a bug that made it hard to use: where commenters enter their name and email, their input would appear with a white font on white background (invisible). (1/5)
I integrated SitePing into the site using a coding agent. After testing I pointed out the invisible type problem I had observed to the agent, and it found the root cause in a minute or two. So I decided to let the agent create an issue upstream.
The issue: https://github.com/NeosiaNexus/SitePing/issues/157 (with an additional comment about the root cause and a suggested fix). (2/5)
Description With theme: "auto" and the operating system set to dark mode, the name/email inputs in the identity modal become white text on a white background as soon as they are focused, making the...
A few days later, the issue got fixed, likely by the maintainer's coding agent, confirming the root cause analysis and the suggested fix in the issue created by my agent.
PR: https://github.com/NeosiaNexus/SitePing/pull/158
(3/5)
Summary Fixes #157. With theme: "auto" resolved to the dark palette (OS dark mode), interactive surfaces that hardcoded background: #fff turned white-on-white because the text color stays...
So for Open Source projects, I understand that this is blessing and curse at the same time, as the load for maintainers is definitely increasing with more incoming issues and PRs.
My hope is that the average quality of incoming bug reports at least rises if AI-assisted. If you're maintaining or observing a project and have some experience, please let me know. (5/5)
Marian