No you don't. Feed the essay to Claude and ask it to summarize it for you, or just use the Jira MCP and have Claude or codex take the ticket, use the GitHub MCP to get it the source, have it go work on the ticket, have it generate the code, generate some unit tests, then you go literally yell at your computer using Wispr Flow or some other transcription software to tell Claude how to fix the mess it made, and then you after you've cleaned it up, you submit the PR.

When ChatGPT first came out three years ago, we joked about Devin and having an AI coworker, but I was just given 5 tickets to work on before break, and damned if AI didn't take a well scoped ticket and just did it before I even finished reading that very tightly scoped ticket.

While I personally find that generative AI has helped me be more productive and even boosted my ability to learn things, I really _dislike_ that we’ve normalized this behavior:

Whether a Jira ticket, an email, a yearly review, we feed bullet points into a black box to get a bunch of fluffy text. On the other end, we feed the fluffy text into the black box to get bullet points.

We’re killing penguins because we’re somehow afraid to just send the simplified bullet points to each other in the first place.