I'm quite surprised at Windows Recall's reception; it's almost a visceral horror at the idea.

I say this as someone who is very thorough with respect to security and privacy. For example, I only do online banking on a virtual machine. I'll think about any web search and ask the question "Do I want this in my history?" and use a privacy-protecting search if not (for example, anything medical or health related doesn't go through a signed-in session). I've never trusted anything like Mint, and for years have downloaded CSV transactions - parsing, categorizing, and graphing things myself in Python.

I'm also well organized. I build information hierarchies and robust systems of personal and professional notes. Years ago, I ran various software for personal wikis (my very first pull request was to a git-based wiki Gollum in 2012 as one iteration of that system). Now it's trees of markdown files. I have mail archiving/tagging/filtering rules so long I have to split the match field into two different rules.

With that context established, as someone both well organized and very much on the secure/private end of the spectrum, I think Recall would be really freaking useful. I absolutely want a well implemented, secure version of this. We're finally at the point where a large class of new interactions with computers are possible, the sort of things folks have been imaging for decades. Here's a 2008 story about Gordon Bell attempting to capture every moment of his life for a decade.

Bell has gone on to collect images of every Web page he has ever visited and television shows he has watched. He has also recorded phone conversations, images and audio from conference sessions, and with his e-mail and instant messages.

And this bit is just amazingly prescient:

In 20 years, digitizing our memories will be standard procedure, according to Bell. "It's my supplemental memory and brain," he noted. "It's one of my most valuable possessions. I look at this thing and think, 'My whole life is there.'"

These ideas aren't new; it's just that they're finally becoming readily possible. I remember seeing Rewind.ai last year, thinking it sounded extremely useful, and ruling it out because it doesn't process fully locally. Just like I remember seeing Mint locally years ago, thinking it would be quite useful, and rejecting it in favor of writing local Python: it's just too much to trust off-device with where the industry's security is now (and will be for any foreseeable future).

I do think a secure implementation is possible, but it looks like Windows didn't even try from Reading Kevin Beaumont's great post (and turning it on by default was just 🤦). What I don't agree with, though, and what seems to be a broadly shared sentiment, is:

In practice, that audience's needs are a very small (tiny, in fact) portion of Windows userbase - and frankly talking about screenshotting the things people in the real world, not executive world, is basically like punching customers in the face.

Really? I would love to be able to ask "What was the PR I left a comment on asking for several changes, but approved?" or "What are the cocktail recipes I was looking at last month with Yellow Chartreuse?" There are just so many things that are gone and quite hard to find if not explicitly captured in the moment. Even in just a purely personal context, I have folks I literally chat with in 4+ channels (Instagram DMs, texts, email, and Google Chat - and that's just one person). Every platform is trying to capture its users inside the boundaries of their own walled garden.

Aggregating our digital information in a way that's queryable via natural language is one of the most useful things I can imagine. An implementation with great defaults (the preview they shipped didn't even exclude Chrome private windows, just Edge... seriously?), robust user-controlled rules, TTL, commands like 'erase the last 5 minutes,' and much better-protected local data would be extremely useful.

Maybe this can't ever be done securely, but new modes of interacting with and making use of the ever-growing caches of digital information we generate are just becoming possible in a way that I think folks are going to find very, very useful.