edje
Log in

Training · Course catalog

Intermediate AI: Hands-On AI Agents with Claude Code

Intermediate AI is a full-day, in-person course from edje (Seattle/PNW) that turns AI chat users into supervisors of autonomous AI agents. Participants operate Claude Code on their own laptops, climb a graduated autonomy ladder, and complete two measured capstones — a 90%+-accuracy text classifier and a working personal CRM application.

Bring this to your team← All training

At a glance

Intermediate
level · prerequisites: Intro to AI: Practical AI for Professionals, or equivalent working familiarity with an AI assistant
1 day
In person, full day (8:30a–4:45p)
8.0
NASBA CPE credits (Group Live)
4 weeks
of office hours included after the course

About this course

A hands-on working session in which professionals move from AI chat users to supervisors of autonomous AI agents. Working in Claude Code on their own laptops, participants install and operate a terminal-based AI agent, write steering files (CLAUDE.md) that encode their working rules, put real folders under git version control using plain-English commands, and climb a graduated “autonomy ladder” — permission allowlists, sandboxes, and a machine-reviewed auto mode — that trades babysitting for engineered safeguards. They enforce hard rules with event hooks, practice plan-first supervision by editing build plans before any work executes, and finish with two measured capstones: reverse-engineering a text classifier from labeled data to 90%+ tested accuracy, and planning, building, versioning, and extending a working personal CRM application. Governance principles from Day 1 — least privilege, reversibility, containment, human oversight — are applied at every rung.

Instructor: Aaron Sparks, Founder, edje. Delivered in person, on-site at your office in Seattle and throughout the Pacific Northwest.

What you'll learn

  • Pick the right AI surface (chat, document workspace, desktop agent, terminal agent) using audit-trail, measurement, autonomy, and building requirements.
  • Install and operate Claude Code: terminal navigation, session management, slash-command controls.
  • Author CLAUDE.md steering files encoding standing rules — versioning, secrets hygiene, style adherence, approval boundaries.
  • Use git through natural language — commit, diff, log, restore, branch, worktree — to make agent work reversible and auditable.
  • Configure graduated autonomy: allowlists, sandboxes, machine-reviewed auto mode, matched to task risk.
  • Implement event hooks that deterministically enforce rules an agent cannot override.
  • Use plan mode plus a structured requirements interview to specify and edit build plans before execution.
  • Derive a text classifier from labeled ground truth and evaluate to a defensible 90%+ accuracy on a held-out test set.
  • Plan, build, version, and extend a working application by reviewing diffs, not code.
  • Match model tiers (including long-running frontier models) to task complexity and duration.

The day at a glance

Day 1 recap → One Claude, four surfaces → Claude Code installed → Steering files → Git: the undo button for everything → The autonomy ladder → Hooks → Plan before build → Capstone: the classifier → Capstone: your local CRM → Wrap-up.

Built for doing, not watching

17+ engagement elements including 9 hands-on exercises with two extended capstones. Practice datasets provided (75 labeled customer inquiries; starter CRM contact data).

Logistics & what to bring

  • Paid Claude plan (Pro, Max, Team, or Enterprise).
  • Laptop with permission to install signed software and IT pre-approvals; Windows participants install Git for Windows.
  • Pre-training checklist provided on registration.
  • Includes four weeks of weekly office hours after the course.

Frequently asked questions

Do I need to be a programmer?

No — git and the terminal are taught in plain English for non-programmers.

What do I build?

A measured text classifier (90%+ tested accuracy) and a working personal CRM you keep.

Do we earn CPE credits?

Yes — 8.0 NASBA CPE credits (Group Live).

What's required?

A paid Claude plan and a laptop you can install software on. Prerequisites: Intro to AI or equivalent working familiarity with an AI assistant.

Also in the catalog

Intro to AI: Practical AI for Professionals

From installing an AI assistant to running your first working automation in one day — prompting, guardrails, connectors, and a four-job capstone. No technical background required.

View course details →

Bring this course to your team

Tell us about your team — size, roles, and which AI platform you use (or want to). We'll reply within one business day.