Nave Ben Naim

Founding Engineer at Naboo.ai. Security research, CTF writeups, and notes on AI context engineering.

The Five Walls of a Context Layer

Building your own context layer for AI agents is harder than it looks - starting with the one wall nobody sees first.

Posted on September 15, 2026

I started a new series on Medium: The Five Walls of a Context Layer. [Read More]

Tags: AI Context Naboo Knowledge Graph Ontology LLM

How to build the right searching tools?

Four search methods, one incident, and why reaching for embeddings first is usually wrong.

Posted on August 28, 2026

Everyone reaches for embeddings first. It is usually the wrong call, and it costs you tokens and relevance. [Read More]

Tags: Search RAG Retrieval Embeddings AI Learn

Make your codebase teach AI agents how to work in it

A guardrail passed while checking the wrong code. Fixing it taught us what an AI-native codebase requires.

Posted on August 28, 2026

An agent opening your repo for the first time is a new hire on day one, except it forgets everything by tomorrow. Give it nothing and you get the same three problems every session: [Read More]

Tags: AI Native Agents Codebase RFC Claude Learn

What it took to turn a coding agent into an R&D teammate

Building an agent that could remember team decisions, investigate problems, and ship reviewed code.

Posted on August 28, 2026

During a one-hour customer meeting, we heard seven complaints and feature requests. We passed each one to the AI Harness teammate our R&D team had built. It worked on them concurrently, opened seven pull requests, and ran end-to-end tests before the meeting ended. [Read More]

Tags: Harness Agents AI LLM Learn

A Smart Model Doesn't Make Up for Bad Context

With the right context, a lean and cheap model can give you the results of an expensive one.

Posted on June 21, 2026

A smart model doesn’t make up for bad context. [Read More]

Tags: AI Context Naboo LLM Claude

Older Posts →