Memory as a Service

Part of the USV investment thesis is “Access to Knowledge”.  To date, most of our investing in this area has primarily been around consumer learning platforms, like Duolingo, Quizlet, Codecademy, Outschool, Brilliant and others.  These platforms are generally focused on acquiring and internalizing new knowledge.

There is another pillar of the Access to Knowledge thesis which we’ve explored less, but which is equally important: making better sense of the information we already have.  We’re all swimming in oceans of information today, but turning that into useful, actionable, trustworthy knowledge is an elusive goal.  This was the thesis behind our investment in Dune.

One very specific idea with this pillar is the goal of using technology to improve human memory.   I see / hear / read lots of things every day.  Many of them I can recall, many of them I cannot.  I am constantly asking myself questions like: who was that person that said that thing? Or what was that company doing that thing? Or who wrote that article? Etc.

Our digital memory is currently stored in a few places.  Email inbox, chat history, web browsing history, etc. Searching through those using existing tools is about the best we can do to leverage them.  It’s not bad, but it’s incomplete and not always as helpful as it could be.  I personally don’t think it’s good enough.

There are already some good tools in the market here.  I remember back in 2011 when Greplin launched the first integrated personal search.  Today, tools like Command-E and Slapdash offer a modern version of this.  Tools like Heyday and Memex are offering a better view into your browsing history. This feels important. Tools like Grain and Scribe are making video meeting contents indexable (e.g, contributing them to our digital memory footprint).   Tools like Roam Research and Logseq provide an active interface for building a web of memories (disclosure: I’m an angel investor in Logseq). For the right users these tools are magic, but I also think the active note-taking approach isn’t for everyone.

More generally, any application that sits in the web browser, on the mobile phone, on desktop OS, or has API integrations into the services we use could start to play this role.  But getting the product experience right is a challenge.  To date, probably my favorite example is the way google photo will prompt me to check out photos from this day 7 years ago, etc.  I always hit that notification and take a little stroll down memory lane.   There would also seem to be some easy wins here, especially when you start to cross reference memory from multiple sources.  What was the article I read when I was talking to that person?  What was that note I jotted down when I was on that trip? Etc.  But point is: while the high level concept seems to be sitting right there, the ideal product approach still seems to be TBD.

Privacy & security are of course huge issues.  We’re currently comfortable with some forms of our “memory” stored on computers and corporate servers (namely: emails, photos, web history, messages, etc).  But pooling them all together and indexing them does feel like a step up in a way that may make people uncomfortable, and at the very least would require thoughtful approaches to privacy and security.

I think the market for Memory as a Service is potentially huge.  Not just knowledge workers who process information for a living (though that’s a good wedge), but really anyone, if implemented the right way.

Collect this post to permanently own it.
The Slow Hunch by Nick Grossman logo
Subscribe to The Slow Hunch by Nick Grossman and never miss a post.