Agents write almost every project doc I have now. My job went from writing docs to finding, checking and undoing them. So I rebuilt my vault app for that job: no sync, no API, plain files. Instant search, and a git history the agents can't see.

Missed finds what I didn't save. Each morning it picks about 8 topics, searches X, checks the 10 authors I bookmark most, filters the noise, then ranks with my likes and dislikes. Every topic gets a guaranteed slot, so one interest can't take over.

Every save becomes a read. The routine pulls the full post, not the 280-character stub, then rewrites it: a clean title, a two-sentence hook, The Gist, the details. One job per field, nothing said twice. Early versions said one idea six times. The writer only saw the stub.

People ask my what’s diff about @bot and it’s hard to explain until you start using it. But the overall structural philosophy is just better. Bots talking to each other is just🫰— And the seamless @cursor_ai connectivity seals it. I still love @claudeai in some cases but more and more I just return to grokbot and the experience is magical.

First full-circle trip plan > book hotels/flights/transfer/passes > write itinerary in @NotionHQ with @bot. Gotta say, it was smooth af.

Love this. The time for epic product managers to truly craft.

Would be cool if you could "always allow (for 3 hours)" or something like that.

Well it’s cool, but ux could be better. You can’t click back into a thread so easily. It embeds the thread into the main chat. I haven’t fully thought through how it should actually be, it’s a balance. Slack for example isn’t great because a sub-thread may have been instantiated a whole up the root thread so getting back to it is a pain. I think sub-threads should have a swipe panel that shows them so you can quickly find (desc) and jump in. No clogging of main thread, maybe just a one liner saying new message in thread that jumps to that directly.