Why speed feels great but structure keeps you sane
Every day I see developers firing off prompts to tools like Cursor, Claude Code, and ...
For further actions, you may consider blocking this person and/or reporting abuse
Nice article. I've written two here that are also trying to help solve this dilemma. I've been concentrating more on keeping a durable memory of what I've decided with the AI.
A habit I'm developing is:
I'm a big fan of AI's writing docs for AI's to read making it easy for their context windows to catch up to where you left off.
I second that.
AI makes it easy to keep your documentation up to date and comprehensive, which in turn makes it more valuable for agents (as kind of long term memory) and humans (yes, humans love documentation as well).
At the end of a session, I ask the agent to conduct a retrospective and store it in my current spec directory, and amend the lessons learned in the contributors' documentation, as well as to have a look at the documentation in general and update it with the current work.
Thanks.
This resonates with me, as I took a similar approach in my project.
I used GitHub Speckit which provides most of the prompt boilerplate, so you don't have to get deep into prompt engineering.
Interesting article, it's good to use AI as an assistant and not depend on it 👍🏼
That's the idea everyone have to have in long run
Thanks for sharing
Absolutely love the balance you’ve struck between speed and structure!
A very important topic—how to work effectively with Code Vibe. I once heard a great definition: every developer is essentially the team lead of their LLM. You give the instructions, set the expectations, and you're fully responsible for everything the system produces.
I wrote a post about this:
dev.to/sara_h/im-a-junior-but-im-a...
Nice article.