Mitali Gupta

Soo, Claude is responsible for my trust issues now

Getting falsely accused of using AI is starting to feel like this generation’s witch trial🧙‍♀️ More and more, it feels like an AI detector is enough to make people question someone’s work. The hard part is that once you’re accused, you’re expected to prove you didn’t use AI. That’s not an easy thing to do. I’ve heard stories of writers losing freelance work, students struggling to defend their assignments, and employees being questioned over emails they genuinely wrote. AI has incredible potential, but we also need to be thoughtful about how we use tools that can make mistakes. Has this happened to you or someone you know?

The first 30 minutes you spend setting up Claude are probably worth more than the next 300 prompts you write. When I first started using Claude Code, I kept writing longer and longer prompts, hoping it'd eventually "get" how I work. It didn't. Not because Claude isn't good. It just didn't know me. It had no context, no idea what I was building, how I like to work, or the rules I expect it to follow. So instead of writing better prompts, I spent 30 minutes setting up my environment. First, I created a shared context folder that Claude can reference whenever I start a new project. Then I created three simple files: 👉 what I do and the kinds of projects I work on 👉 how I like responses to be written (and the phrases I never want to see) 👉 my working preferences and non-negotiables Finally, I added a few global instructions: 👉 always show me a plan before making changes 👉 ask before executing destructive actions 👉 explain trade-offs when there are multiple approaches That one-time setup changed the way I use Claude. Now, instead of spending half my prompt explaining myself, I can jump straight into the problem I'm trying to solve. It's one of the highest-leverage things I've done with AI this year. I'm putting together the exact folder structure, files, and instructions I use. If you'd find that useful, let me know and I'll share them. #ClaudeCode #ContextEngineering #AIWorkflows #GenAI #BuildInPublic

So.. Job search's been fun🫠

I don’t know who needs to hear this, but waiting for the prompt and the output to complete their enemies-to-lovers arc is a perfectly acceptable reason to stay up until 3am 💁🏽‍♀️ What’s the last bug that kept you up all night?