AI as we know it today is unlike any technology shift I’ve lived through. I’ve tried to stay genuinely skeptical without letting that skepticism kill curiosity–so I’ve been pushing myself to keep exploring. Here’s what I’ve found, as of April 2026.

Where it’s better than ever before

• One-off feedback loops like I wrote about in How I Used AI to Audit My Job Search (And You Can Too)
• Repeatable feedback loops like Marshall’s reflect skill for Obsidian
• Taking my messy notes and drafting them into a linear note I can build from (and rewrite in my own words–an important lesson in not publishing slop)
• Low risk coding i.e. hacking on my website

When the personal cost is higher than the value

• Work that replaces habits that help me learn. Hazel Weakly wrote a beautiful piece on learning patterns and where not to use AI, and I found it profound in part because I’d already failed my way to the same conclusion. I was having such a good time taking old notes and getting AI to reframe them as blog posts that I lost the thread on why I share anything publicly. Writing is a reinforcement pattern–an opportunity to sit with what I’ve learned until I get clarity on the story I tell myself. Asking an agent to clean up my drafts and publish them was a fast lesson in slop.
• Tasks that require collaboration I try to rush past. I once used AI to migrate software documentation for a team that had low investment in their docs. The problem wasn’t the tool–it was that I hadn’t done the work of clarifying the product taxonomy, customer needs, or shared definitions before trying to ship. My “I did this in two hours!” turned into a stale PR that just confused people. Takeaway: bring the right people along from the beginning if you want it to stick.

What I want to explore

• Unapologetically vibe coding where I’m a subject matter expert but low on time. I’ve noticed a perfectionism trap: waiting until an idea is great before giving it a try. I want to fail in public more often, and AI is sure to power some of it.
• Improve my leadership and delegation skills by learning to articulate what I need to multi-agent workflows. I love managing people–some who mostly need me to get out of their way, others who need a lot of coaching–but I’ve found I still bifurcate “my work” from “their work” sharply. I want to get more effective at delegation, which means getting better at describing what I’m after and how I’d know we got there. Multi-agent workflows look like a way to practice this.

My next exploration to this end is exactly why I’m working on exploring-multi-agent-ai-workflows. I’d love to hear what you’re exploring.