Why Obserf never posts

Obserf reads the thread, ranks it, and writes the comment. The last step — pressing send — it leaves to you. That looks like an unfinished feature. It is the design.

The asymmetry

A missed opportunity costs one comment nobody read. A wrong comment — posted into a thread the model misread, or a venue that bans self-promotion — costs the project’s standing in the exact community it was trying to reach, permanently and publicly, under the maintainer’s own name.

The upside is bounded. The downside is not. That is enough to decide it.

What the model cannot know

An opportunity has to be three things: useful to the reader, permitted by the venue, and free. Obserf assesses the first. Whether a mention helps is legible in the thread itself.

The other two are facts about the venue’s rules, and the model has no tools — it cannot open a subreddit’s sidebar or a directory’s submission page. It uses supplied evidence, including any verified venue rules you add to the project profile, and treats missing evidence as uncertainty. Silence in a search snippet is not proof that promotion is allowed, or that a listing is free.

What reaches the inbox is not known to be forbidden and not known to cost money. That is a weaker claim than permitted and free, deliberately. Closing the gap means reading the rules, which is a person’s job.

It is structural, not a setting

There is no posting code path in Obserf. Not a disabled one, not a flag. Posting would require adding a new code path, not changing a setting.

It is not a credential boundary, and the difference matters: the gh token Obserf borrows for GitHub search may well be able to write. What guarantees Obserf does not write is that nothing in Obserf asks it to.

The same reasoning shapes the drafts. Every one has to be useful with the link removed, and has to disclose the affiliation in your own voice. A draft that only works when nobody notices who wrote it is one you should not be posting either.