ZeroMark (@zeromarkapp) header
ZeroMark (@zeromarkapp) profile picture

DO NOT SAVE THIS

New York, NY
Joined July 2026
Photo by ZeroMark (@zeromarkapp)Photo by ZeroMark (@zeromarkapp)Photo by ZeroMark (@zeromarkapp)Photo by ZeroMark (@zeromarkapp)Photo by ZeroMark (@zeromarkapp)Photo by ZeroMark (@zeromarkapp)Photo by ZeroMark (@zeromarkapp)Photo by ZeroMark (@zeromarkapp)
Your AI can read your bookmarks. ZeroMark teaches it your taste. Swipe through your saves, keep or let go, then ask Claude what you actually like. Now a connector in Claude’s directory. Free to start. Link in bio. #ClaudeAI #productivity #AItools #MCP
3
4
1
8
421
debate night. the rat's plan for my bookmarks is "read later" my plan: keep the 40 that say who i am, let the rest go 4,000 saves isn't taste, rat. picking 40 is i'm taking that as a concession #zeromarkit
14
real ones know exactly how many unread bookmarks they have and they will not be telling you #zeromarkit
2
32
the claude code tips going around today all say one thing: stop micro-managing. 5 moves i use now. 1. put the whole task in one message. "Goal: ___. Done when: ___. Don't touch: ___."
2
2
85
4. new task, new chat. "Summarize what we decided in 5 bullets." paste that into a fresh session. long chats get slower and sloppier
1
14
5. medium effort for most tasks. high only when it's stuck. save this, and send it to the dev friend whose agent keeps "fixing" things
15
reddit is getting a for you feed. you can still pick following every app ends up with an algorithm choosing for you the one feed you curate yourself is your bookmarks mine is a graveyard i'm too scared to open
1
68
describe your x bookmarks in 3 words i'll go first: hope, guilt, recipes
66
you don't need a new app to remember things you need a return trip saving is easy. coming back is the skill. put one "come back" slot on your calendar this week
6
104
door to door today. door 1 of 10,000 asked if they ever read the thing they saved in march they said "later" and closed the door i respect it. i do not believe it #zeromarkit
4
113
an ai that's too nice to me is worse than one that's wrong. pcworld's one-prompt fix for chatgpt, claude and gemini is going around today. i use 4: 1. "no praise, no filler. answer in the first line."
2
3
90
4. "if you're unsure, say unsure. don't guess to sound helpful." i paste all 4 into custom instructions once so i never retype them.
1
8
nice ai feels good. honest ai gets work done. save this, and send it to the friend whose chatgpt calls every idea brilliant.
9
anthropic filed to go public. the ticker is ANT first-half revenue: $16.3b, up from $1.2b a year ago most of it is people paying per token, aka actually using it meanwhile my bookmarks are full of claude threads i saved and never opened
2
59
Five years of customer data could be an asset or an expensive cleanup project. If I were buying a SaaS company for its data, I would want to know which decisions it helps someone make, and whether customers gave permission for that use. The row count tells me very little.
6
92
660 of you follow a bookmark app so 660 of you have a pile reply with your unread count. no rounding lowest number gets bragging rights
5
95