Essays & Deep-Dives

The Yohanun Blog

A small number of essays we actually stand behind: how to think about LLMs, why memory needs an architecture, and why governance is the layer everyone skipped.

Featured

Product New

Introducing Yohanun Guilds

Every coding agent starts every session from zero. Guilds gives each of your projects a named agent with memory of its own, and the difference is the first thing it says.

Read Full Article

claude · session start

Setanta — guild memory loaded: 3 open tasks, last worked yesterday

I'm Setanta, this project's agent, carrying the memory of every session we've had here.

Where we left off:

· Programme visibility shipped and verified on prod

Essays

Field notes from building the platform, not a content calendar

Perspective New

Memory Is an Institution, Not a Database

Whose version wins, who may see it, does it still stand: the hard problems of AI memory were solved centuries ago by courts, libraries, and one small town in Donegal. On building memory by translating institutions.

We Unplugged Our Agent's Memory. Here's What Changed.

The same Claude Code sessions run twice on our own production repo, once with guild memory and once without. Real transcripts, honest ties, and the three kinds of knowledge that never reach a repository.

We Ran mem0's Benchmark Against Ourselves. Here's Everything.

Out of the box versus out of the box: 62.4% to 48.8% under one strict judge, with the full method, the fairness ledger, and the null results published alongside the wins.

Building AI with Memory: Beyond the 50 First Dates Problem

The essay that started Yohanun: why every conversation with AI feels like starting over, and how a memory runtime solves the problem that bigger models never will.

How We Learned to Think of LLMs as Compute Engines

The biggest architectural mistake teams make with AI isn't technical; it's conceptual. Why treating LLMs as specialized compute, not complete systems, changes everything you build.

MVC, AI Systems, and the Memory Problem

Where traditional application patterns work well for AI, where they break down, and why intelligent systems need an infrastructure layer MVC never anticipated.

Want to argue with any of this?

These essays are opinions held with reasons. If you're building in this space and see it differently, or the same, we'd genuinely like to hear it.

Email Us →