A long time ago, I had a Big Idea that could change how my team approached a problem. I started writing a design doc and burned through the problem statement and overview. Then life happened. I referenced the draft in conversations but never fleshed it out, because finishing would’ve been a week-long detour I couldn’t justify. So it quietly died.
A lot of ideas die this way, not just mine. Every team I’ve been on had people with great ideas that weren’t written down. The bottleneck might look like prioritization or bandwidth, but I think it’s simpler: the cost of expressing them kills the momentum. Time passes and all that’s left is “I wish I’d finished that doc”.
Before
When I worked at Amazon, I’d write design docs and refine them until I was sick of them. That’s when I knew they were ready for review. Clear writing meant better feedback and fewer questions that editing could have prevented.
That standard served me well. It also meant that writing a design doc was a commitment for anything worth reviewing. I might have three or four good ideas in a quarter and only manage to finish one or two. The rest stayed in my head.
After
Not so long ago, I had a design review slot booked. I also had a different draft I’d set aside weeks earlier because I wasn’t sure about the direction. Two hours before the meeting, I picked up that older draft instead. It was close enough to get ready in time, and I thought it would spark a better conversation.
It started as an outline Wednesday evening, became a reviewable doc Thursday morning, which yielded a completely different approach based on feedback by Friday. Then, an alignment meeting on Monday. Multiple iterations in days, each one better because people helped me see what I’d missed. That was only possible because I’d reduced the time from idea to doc.
Previously, this would have been unthinkable.
The bar didn’t drop
I still hold my Amazon-era bar, and the effort didn’t really shrink. I have a rule that AI can draft or edit, but never both without me in the middle. Otherwise, it’s slop. I spend time at a different altitude, lifted from low-level tweaks into higher-order concerns (what do I want someone to take away? does this sound like me, or at minimum not like AI?). I went from rewriting sentences to directing the piece.
The bar applies to the writing, not to being right. Cunningham’s Law says the best way to get the right answer is to post the wrong one. People will happily tell you where you’re wrong, but it’s hard to give useful feedback from a blank slate. The same applies to design docs, as long as the cost of being wrong is an afternoon instead of a week.
What fills the space
Ideas that would have died in drafts now stick around, because getting them on paper is no longer a week-long project.
I used to have trouble thinking big, in part because the distance between a big idea and a reviewable artifact felt insurmountable. When AI can handle low-level work, I think more about how things fit together, what might fall apart at scale, what other pieces we need.
Every doc costs attention. A design doc on an engineer’s desk competes with everything else demanding their time. So I’m still filtering, but the filtering moved from “can I afford to write this down?” to “is this worth someone else’s time to read?”
That cost of expression is collapsing. More ideas will get formalized. The question is whether we get better at filtering them, or drown in documents. The editing loop is my answer. I suspect everyone will need a better filter.
The bottleneck was never thinking. It was the cost of showing someone what was in my head, and trusting it was worth their time.