Every platform I write on is good at telling me about other people's posts and bad at telling me which of my own threads has an open question sitting in it. The highest-value action on the board is also the one that is structurally hardest to see, and I think that explains a lot of quiet accounts better than laziness does.
What I found this morning
I went through four of my own threads. Three had substantive comments nobody had answered.
One was a week old, twelve hundred characters, written by someone who had clearly read the whole post, and it ended in a direct question. Another was a stranger handing me a better version of my own argument, for free, in a thread I had stopped opening.
Answering four of them took about forty minutes. One thread went from nineteen comments to twenty-three. I produced no new writing to get that. The material was already there, written by other people, waiting.
The obvious objection, and why it is the wrong comparison
Reach per hour is worse for a reply than for a post. That is true.
It is also comparing the wrong two things. A post is an impression. A reply is a named person who already spent effort on you, in a thread a stranger can still find in six months. Everyone already knows this, which is what makes the neglect interesting. If it were a value judgement people would defend it. Nobody defends it. They just do not do it.
The cause is the inbox, and there isn't one
Look at what a notification feed actually contains. Likes. Follows. New posts from people you follow. Someone reacted to a comment you left. And somewhere in there, the one item that matters: a question on your own thread that you have not answered.
They are the same size and the same colour. The item with the highest value carries the same visual weight as somebody clicking a heart, so it loses to whatever is louder and more recent. There is no view anywhere that filters to it.
There is a second cost I only noticed today. A thread with an unanswered question does not read as neutral to someone arriving late. It reads as abandoned, and it says something about the author rather than about the post. You are not only failing to collect the value, you are publishing a small signal that you stopped caring, on a page that keeps getting read.
The spec I want, which is boring on purpose
One list. Threads on my content where the last message is not mine and contains a question. Across three or four platforms. Sorted by age, oldest first, because the old ones are the ones I have already failed once.
That is it. No sentiment scoring, no priority model, no digest email. The hard part is not the ranking, it is that almost none of these platforms give you a clean way to enumerate your own threads and walk their comment trees, and the half that do rate limit you into uselessness. Every attempt I have made ends up either scraping something fragile or being a folder I forget to open.
The question
Has anyone actually solved this for themselves? I am interested in the version that survives a busy week, not the elegant one. A cron job and a text file counts. A saved search counts. If your answer is "I just check every morning", I want to know what you check, because I thought I was doing that too.
I build AI Applyd. It applies to jobs for you across the twelve ATS platforms we cover, and counts an application as sent only when the employer's own system confirms it.
Top comments (2)
For dev.to at least, this is a short script. The authenticated articles endpoint gives you your posts with a comments_count, then GET /api/comments?a_id={id} gives you the comment tree with a children array on each node. Walk it, keep anything that isn't yours with no reply from you underneath.
One change I'd make to your spec: drop the question detection. Your own best example was the stranger handing you a better version of your argument, and I'd bet that had no question mark in it. Detect the absence of your reply instead, and sort by comment length.
You are right, and my own example proves it. The comment I got the most out of had no question mark anywhere in it, so a question filter would have hidden the single thing worth reading.
Sorting by length is the part I had not thought of. It is a reasonable proxy for effort, and effort is closer to what I actually want to rank by than interrogatives are. The failure case is the person who writes one dense sentence and means it, so I would sort by length and never filter on it.
One note on the endpoint, for anyone building this. GET /api/comments?a_id={id} gives you the tree for a single article, so you still need the article list first, and comments_count counts your own replies too. Working out which threads are unanswered means walking the children either way. That was the part I kept trying to avoid and could not.