A while ago GitHub informed me that they would sponsor Copilot for our open source project AlaSQL (a JavaScript SQL database for the client).
I had never really thought much about it. I actively try to ignore the industry push to shove AI into everything. But a few days ago, I somewhat accidentally assigned Copilot to an issue. To my surprise, a nice PR was created shortly after with a spot-on solution and a test file. A one-shot fix!
Impressed, I went to our oldest issue (11 years old). Click. Assigned. Let's see.
After 48 minutes, I was asked to review the PR. It had one-shot fixed an age-old problem with `ORDER BY` across 3+ `UNION` clauses. This involved updating the grammar for the lexical analysis and inserting the right controls to ensure the ordering executed at the right time. Nice.
I immediately went slightly rampant on old (and well-described) issues, and after a combined engagement of 5 hours across 2 days "herding" the AI, the next version of the AlaSQL library was released - fixing issues with a combined age of 109 years, 1 month, and 28 days.
Sounds like the real complaint here is not about Jira but about people using Jira wrong, because they dont understand how you facilitate a productive environment for the lifecycle of the delivery across product, design, QA and developers of different stacks.
I love this project! But I feel like the attention to details in planning the project is not matched with the attention to details when it comes to capturing the essence / energy of the font as the tree progressed. Some examples are "a" for year 3 and "h" for both year 2 and 3 where the font is lacking a lot of the expression that the designer tree is providing.
But very interesting. Thank you for creating this.
I had never really thought much about it. I actively try to ignore the industry push to shove AI into everything. But a few days ago, I somewhat accidentally assigned Copilot to an issue. To my surprise, a nice PR was created shortly after with a spot-on solution and a test file. A one-shot fix!
Impressed, I went to our oldest issue (11 years old). Click. Assigned. Let's see.
After 48 minutes, I was asked to review the PR. It had one-shot fixed an age-old problem with `ORDER BY` across 3+ `UNION` clauses. This involved updating the grammar for the lexical analysis and inserting the right controls to ensure the ordering executed at the right time. Nice.
I immediately went slightly rampant on old (and well-described) issues, and after a combined engagement of 5 hours across 2 days "herding" the AI, the next version of the AlaSQL library was released - fixing issues with a combined age of 109 years, 1 month, and 28 days.
Not bad for guiding an agent for a few hours.