Overcomplicate things
This morning I saw a company called Mem0 and I immediately messaged my friend Raf — “Look, this sounds exactly like what we built last year with ImpliedAI!” Rafal responded simply, coldly: “They’re doing even less - just providing a memory api for RAGs.”
Raf and I saw a problem, solved a problem, and continued onwards trying to figure out what consumer application would benefit the most from the solution to the problem. The guys who did mem0 looked at the same problem, and turned it into “open source memory for AI agents.”
I don’t even know what that sentence means. The technology layer they are providing should be implementable by someone with a half a brain in a weekend. But apparently, that’s enough to start a company and to raise tens of millions of dollars these days.
While we looked at a problem and continued looking up to the next problem that emerges from it, someone look down and realized that thousands of people will run into the same problem, and either won’t be able to solve it, or would prefer an existing solution.
This had literally *never* occurred to us.
Look, I am not gloating about being smart here. Although Raf was pretty much top of his class in Engineering Science, the hardest program at UofT. I am the dummy one in this group. But, my point is quite the opposite — our ability to understand the thing got in the way, big time.
I might not be the right person to do this company. Reselling the same 3 functions while pretending it’s the holy grail of the AI future isn’t exactly my cupt of tea. But with that in mind, this is a note to all my engineering-minded readers, you really don’t have to think too hard.
In fact, just stop thinking.
Just because you understand something, there’s a very, very high chance that most people around you don’t. Do with that what you will.
-Kirill.
p.s. If you want to read the post from last year —
