Someone Wants to Build Agent Memory in Zig and Erlang. The Stack Choice Says It All.

A developer posted to Hacker News this week looking for a co-founder to build agent memory infrastructure using Zig and Erlang. There's no product, no company name, and few details — but the stack choice is the point.

Most agent memory tooling today — Mem0, Zep, Letta — sits on Python and general-purpose vector or relational databases. That's fine until you're running agents at any real scale, at which point the latency and operational overhead become a genuine problem. The implicit argument in choosing Zig (C-level performance, compile-time control) and Erlang (the Actor model, OTP, decades of distributed systems hardening) is that the category needs a purpose-built storage and concurrency layer, not another abstraction on top of Postgres.

Whether this particular search finds its partner is almost beside the point. It's a signal that some builders are looking past the framework layer and deciding the infrastructure underneath needs to be rebuilt properly — at the language level, not patched with libraries.

Too early to cover as a product story. Worth a bookmark.