Agentic AI
Course Redesign

Making a developer-oriented course more approachable without losing the concepts that make it useful.

UX / CONTENT STRATEGY / AGENT DIRECTION

VIEW THE COURSE
BEFOREAFTER
01 / BRIEF

THE PROJECT

A UX redesign built with the same agentic tools the course teaches.

The original course offered a practical introduction to agents, tools, context, and web-building workflows. It worked—but it assumed the learner already felt at home around developer language and interfaces.

I redesigned the course for true beginners and non-coders. My role combined UX and content strategy, visual design, instructional thinking, and direction of Codex as an analysis and implementation partner.

ROLE
UX, content, visual design, agent direction
TOOLS
Codex, HTML, CSS, JavaScript, browser QA
OUTPUT
Responsive course redesign + documented system
02 / SYSTEM

What changed—and why it mattered

01

A clearer entry point

Removed redundant lesson teasers and rebuilt the homepage around orientation, module discovery, and a confident place to start.

02

Plain language, real concepts

Reworked definitions, prompts, and lesson framing so beginners could understand agents without flattening the technical ideas.

03

One visual learning system

Unified editorial reading surfaces with dark workbench patterns for simulations, quizzes, traces, and technical examples.

04

Safer agent-assisted iteration

Used scoped changes, protected originals, targeted agent reviews, validation checks, and written handoffs to keep the work reviewable.

03 / EVIDENCE

BEFORE / AFTER

Design decisions you can see.

01

Homepage

From lesson sampler to course entry point

The redesign replaces a passive, centered introduction with an editorial hero, clearer course framing, and a more direct path into the module system.

02

Module page

A stronger reading hierarchy

Typography, spacing, callouts, and technical surfaces were reorganized to help new learners distinguish explanation from action.

03

Simulation

Making the agent loop visible

The simulation was reframed as a focused workbench: clearer steps, stronger contrast, and less decorative noise around the core interaction.

04

Quiz pattern

Feedback that feels part of the course

Quiz components gained a consistent dark treatment, more readable code-heavy prompts, and clearer feedback states.

04 / PROCESS

HUMAN × AGENT

Direction, not delegation without judgment.

01CONSTRAIN

Define scope, protect originals, and tell the agent what it may inspect or change.

02ANALYZE

Use targeted read-only reviews for redundancy, readability, typography, and risk.

03DECIDE

Review findings, reject noise, and choose the design response that serves the learner.

04IMPLEMENT + VERIFY

Make scoped edits, preview in browser, inspect behavior, and document the result.

One handoff documented the next task, constraints, validation steps, and open decisions so work could continue without losing context.

ONE CONCRETE EXAMPLE ↓

AGENT TASK

Scan the homepage for content that repeats material already taught in the modules.

AGENT FINDING

Lesson teasers made the homepage behave like a second, shallower version of the course.

MY DECISION

Remove the teasers and refocus the page on orientation, module discovery, and a clear place to begin.

RESULT

A simpler entry path with less cognitive load and a more confident first impression.

05 / DOCUMENT

THE RECEIPTS

The redesign left a system behind.

Change log — decisions, iterations, rejected ideas, and validation notes.
Style guide — typography, color, components, and visual usage rules.
STYLE GUIDE → SKILL

REUSABLE ACROSS PROJECTS

FROM DOCUMENTATION TO A WORKING TOOL

I turned the visual system into a reusable agent skill.

Instead of leaving the style guide as a project-specific reference, I translated its typography, color, spacing, and component rules into instructions an agent could load and apply in future projects. This made the system portable and repeatable—while keeping final design decisions and quality control in human hands.

06 / REFLECTION

WHAT I LEARNED

Agentic tools move quickly.
Judgment still sets the direction.

01Clear constraints make agent-assisted work safer, easier to review, and easier to continue.

02Beginner-friendly design is often an act of subtraction—not simplification for its own sake.

03Local fixes protect a pilot, but repeated patterns should eventually become shared components.

04AI can surface options and problems. The designer still decides what matters to the learner.

THE HONEST NEXT STEP

Test it with actual beginners.

This pilot relied on self-review, browser checks, and targeted AI-assisted analysis. The strongest next phase would be usability sessions with 3–5 nontechnical learners, followed by accessibility testing and further component refactoring.