The Scratchpad You Already Have

The message composer with the Public checkbox clearly unchecked, a private note being typed into it

Public unchecked. This note is going exactly one place: to me.

Part 2 of 7: A Field Guide to Messages

It is 2:04 a.m. The pager went off. I am staring at a dashboard that is the wrong color, I have a coffee I do not remember making, and I am about to start pulling threads on an incident. And the first thing I do, before I touch a single config, is open the message box and type: p95 latency spiked ~01:50, started right after the cache deploy went out. rollback? checking first.

Public unchecked. I hit post. Nobody sees it. That is the point.

The problem with 2 a.m. notes

Here is what actually happens during a firefight if you do not write things down. You try five things in eleven minutes. Three of them half-work. You get interrupted by someone asking "any update?" in a channel, you context-switch to answer, and when you come back you cannot remember whether you already restarted that worker or only meant to. Then the incident resolves, someone asks you to write it up, and your memory of the last ninety minutes is a smear.

I have lived that smear. The fix is stupid and it works: narrate to yourself, in a place with timestamps, that nobody else is reading. Not a doc you have to name and file. Not a channel where every line is performance. A box that costs nothing to type into and forgets nothing.

Why the message box is good at exactly this

Three things make Messages a genuinely good scratchpad, and none of them are features I added on purpose for this. They just fall out of what a message is.

First, it is private by default. I covered this in the last post but it matters most here: the Public checkbox starts unchecked, and unless I flip it, my half-formed 2 a.m. thought is visible to me and me alone. I never have to think about audience. There is no audience.

Second, every note is timestamped for me. I do not type the time. The system stamps createdAt the instant I post, so my stream of notes is automatically a stream of when. When I write the incident review the next morning, I do not reconstruct the timeline from memory. I scroll my own posts and the timeline is already there, to the minute.

Third, posting is one motion. Type, hit post, keep working. No title, no folder, no "is this a doc or a list" decision. During an incident, decision cost is the enemy. The box has almost none.

A vertical stack of short private messages, each with a timestamp a few minutes apart, forming a running incident log

Six posts, eleven minutes. That is the incident timeline, written for me, without me thinking about it.

It is not just for outages

I lead with the pager because it is the most vivid version, but the everyday version is quieter and I use it more. The link I want to read later but not right now. The name of the restaurant someone mentioned. The one-line idea for a blog post that will evaporate if I do not catch it in the next four seconds. The command that finally worked, so future-me does not have to rediscover it.

All of it goes in the same box, private, no filing. And because a pasted URL auto-expands into a preview card and an @mention auto-links, even a throwaway note keeps enough context to make sense to me later. I do not annotate. I just dump.

The honest tradeoff: this is capture, not organization. The scratchpad is a great inbox and a terrible filing cabinet. If a note turns out to matter (a real decision, a real reference, a thing I will need to hand to someone else) it should graduate into a Document or a List row where it can have structure. The scratchpad's job is to make sure the thought survives long enough to graduate. Most thoughts do not need to. They just need to not be lost in the next eleven minutes.

Which raises the obvious question, and it is the whole next post: a pile of private timestamped notes is only useful if you can actually get back to them. So next, I go the other direction in time and read my own timeline backward.

Adron