Re:Thoughts on memory layers like mem0
by anshmalik2002 on Monday November 16, 2026 @07:29AM
(#50516901)
Don't use grok and gemini, but based on your comment, I am surprised that they haven't cracked memory yet.
I actually like Claude code for memory, but you're right from a token consumption perspective, it consumes a lot more tokens, didn't really think much of this before.
I agree that it is a hard problem at scale. But at that scale, companies should be able to build their own memory layer for their agent's use case, instead of relying on a memory layer which is built for a more general use case.
I think there's space for a Heroku/Netlify/Firebase style service. Somewhere between large scale and none.
It took me a few weeks and some talented/passionate engineers to figure out embeddings and vector DBs and searches. And another few weeks to set it up to a production level. Markdowns are cute, but every second of processing time is the difference between people using it and people dropping off.
If you have something that's just plug and play, and fairly cheap to start with, that would help a lot of people. Some would just stay locked in rather than hire an engineer to deal with it.
I do think it's actually a Hard problem, otherwise everyone would have done it. ChatGPT is the king at it, but they're not selling memory as a product, so I assume it's tightly coupled with the chat. Claude can't do it without token guzzling.
And the first thing I set up with AI chatbots is memory. You need memory to work with past conversations, and suddenly you're dealing with emergent behavior and token drain.
The other question is timing.
It's probably a year late. You could easily make a billion dollar company just from this, but just as easily be disrupted by Big AI.
There will probably be a time when this would be unbundled, similar to how telcos and cloud unbundle all kinds of offerings. And that opportunity will come several years later.