TL;DRAI

Hillock: offline memory for local LLMs via SQLite + Hebbian plasticity + 10k-dim hyperdimensional computing, fully open-source Python, no vector DB. Qwen 3 (5.2GB) scored 25% vs Qwen 2 (1.5B) at 50% because larger models extract precise semantics that rigid benchmarks miss—revealing how standard evals fail to capture true LLM capacity.

Hi everyone,

I've been hacking on a personal, local project called Hillock. Honestly, it's very much a work in progress and it isn't some flawless breakthrough, but I wanted to see if we could build a lightweight, completely offline memory layer for local LLMs without the overhead of running a heavy neural vector database or wasting precious VRAM.

It is named after the biological Axon Hillock—the region of a human neuron that sums up incoming electrical charges and decides whether to fire (open the gate) or remain silent (block).

How the architecture works:

The Ground Truth (SQLite): Stores hard facts as simple database triples (Subject-Predicate-Object) so the system has a solid symbolic foundation.

dev.to

I built a local AI memory gate on a CPU, and my 7B model scored worse than my 1.5B model because it was too smart

Hi everyone, I've been hacking on a personal, local project called Hillock. Honestly, it's very much...

lunedì 22 giugno 2026 New tab

TL;DRAI

376 words~2 min read

Hi everyone,

It is named after the biological Axon Hillock—the region of a human neuron that sums up incoming electrical charges and decides whether to fire (open the gate) or remain silent (block).

How the architecture works:

The Ground Truth (SQLite): Stores hard facts as simple database triples (Subject-Predicate-Object) so the system has a solid symbolic foundation.

I built a local AI memory gate on a CPU, and my 7B model scored worse than my 1.5B model because it was too smart

I built a local AI memory gate on a CPU, and my 7B model scored worse than my 1.5B model because it was too smart

Related reading

I built an MCP server that gives AI persistent memory of your SQL database

I'm 18 and I Built a Neuron-Like Memory System for AI

I Built a Local Memory Layer So My AI Tools Stop Forgetting Me

This Half-Gigabyte AI Model Runs Local Agents on Your Phone - Decrypt

Building a Production-Grade MCP Memory Server: Lessons from MindCore

The Zero-Budget Memory Setup Behind My AI Agent Workflow No database. No…

Related reading

I built an MCP server that gives AI persistent memory of your SQL database

I'm 18 and I Built a Neuron-Like Memory System for AI

I Built a Local Memory Layer So My AI Tools Stop Forgetting Me

This Half-Gigabyte AI Model Runs Local Agents on Your Phone - Decrypt

Building a Production-Grade MCP Memory Server: Lessons from MindCore

The Zero-Budget Memory Setup Behind My AI Agent Workflow No database. No…