My Agentic Engineering Setup (2026)

February 2026 • 3 min watch

A quick walkthrough of how I use Beam for day-to-day agentic engineering. This video covers the core workflow: persistent AI agent memory across sessions, one project per agent, and how the terminal setup keeps everything organized.

The Problem: Agents Forget Everything

Every time you start a new Claude Code or Codex session, the agent starts from zero. No knowledge of your architecture, no memory of the conventions you agreed on yesterday, no context about the decisions you've already made. You end up re-explaining the same things over and over.

This is the problem Beam was built to solve.

The Solution: One Project, One Agent, One Memory File

In Beam, every project gets a persistent memory file — a plain Markdown document that lives on disk at ~/ProjectName.md. This file stores your project's architecture, conventions, key decisions, and anything else your agents need to know.

When you start a new agent session, Beam loads the memory file automatically. Your agent picks up right where the last session left off — no re-explaining required.

The Workflow

1. Create a project in Beam — give it a name and a working directory
2. Beam creates the memory file at ~/ProjectName.md
3. Start an AI agent — Beam loads the memory prompt into the session
4. Build, iterate, ship — the agent knows your project inside and out
5. Save your workspace — tabs, splits, and layout persist across restarts
6. Next session — restore the project, start the agent, memory is already loaded

The result: your AI agents compound knowledge over time instead of starting fresh every session. Architecture decisions from last week, coding conventions from last month, bug fixes from yesterday — it's all in the memory file, ready to go.

What's in the Video

• How Beam organizes projects with named tabs and workspaces
• The memory file system and why it works
• Starting an AI agent with persistent context
• The day-to-day workflow for building with agentic engineering