It was late on a Wednesday evening and I was preparing for a big workshop at a client's R&D headquarters. I needed to explain Flight Levels — a framework for coordinating work across different organizational layers — to a room full of managers who had never heard of it.

The usual approach would have been slides. Maybe a diagram in Miro. Something static that I'd click through while people nodded politely and checked their phones.

Instead, I opened Claude and said: "Build me an interactive React explainer for Flight Levels I through III. Users should be able to click on each level and see what it means, how it connects to the others, and what the typical challenges are at each level."

Ninety seconds later, I had a working interactive application. Right there in the browser. No deployment, no coding environment, no Figma mockup. A real, functional tool with clickable layers, animated transitions, and context-specific explanations.

I used it in the workshop the next day. People didn't check their phones.

That was my introduction to what Artifacts can actually do. And I haven't stopped using them since. Most of what follows I picked up from Anthropic's learning platform and then pressure-tested in a year of daily consulting work.

What Artifacts Actually Are

Most people think of Artifacts as "that side panel where Claude shows you code previews." That's like describing a Swiss Army knife as "that thing with the toothpick."

Artifacts are a built-in development environment. When Claude creates one, it appears in a dedicated panel next to your conversation — a live, rendered, interactive piece of content that you can see, use, modify, and share. It can be a React application, an HTML page, an SVG diagram, a Mermaid flowchart, a markdown document, or a full-blown interactive tool.

The key word is interactive. This isn't static output. When Claude builds you a data visualization, you can hover over the bars. When it builds you a calculator, you can enter numbers. When it builds you a presentation viewer, you can click through the slides.

And here's the part that blew my mind: Artifacts can make API calls to Claude itself. You can build an artifact that has its own AI brain. A quiz generator that creates new questions on the fly. A writing coach that gives feedback on text you paste in. A translation tool that adapts to your context. All running inside a tiny app that Claude built for you in seconds.

They call it "Claude in Claude." It sounds like a gimmick until you actually use it.

How I Actually Use Artifacts

Let me walk through the real use cases that have changed my consulting work.

I've built interactive explainers for trimester-based Release Packages, for internal AI developer tooling analyses, for framework comparisons across client engagements. I once asked Claude to build me a Lord of the Rings-themed RPG as an artifact. Full character creation, inventory system, narrative engine. It worked. Was it production-quality? No. Was it impressive? Absolutely.

Artifacts vs. Visualizations vs. Files

Claude recently added inline visualizations — charts, diagrams, and interactive widgets that appear directly in the conversation flow. These are different from artifacts. And both are different from file creation. Here's how they fit together:

The three work together beautifully. A quick inline visualization to explore an idea. An artifact to build the interactive prototype. A downloadable file to deliver the polished result.

Tips for Getting More Out of Artifacts

Be specific about what you want. "Build me a dashboard" will get you a generic dashboard. "Build me a React dashboard that shows cycle time trends over 6 sprints, with a filter for team selection and a comparison view against the organizational average" will get you something useful.

Iterate, don't start over. When an artifact isn't quite right, ask Claude to modify it. "Make the header fixed" or "Add a dark mode toggle" or "Change the chart from bar to line." Claude can modify existing artifacts without rebuilding from scratch.

Use artifacts with your Skills. If you have a brand skill, Claude will apply your brand guidelines to artifacts automatically. My interactive explainers come out in Stellar Work colors because the skill tells Claude which hex codes and fonts to use on dark backgrounds.

Share them. Artifacts can be published and shared via link. Anyone with the link can view them — no Claude account needed. People with accounts can remix them. Incredibly useful for client deliverables or public content.

Don't expect production software. Artifacts are prototypes and tools, not deployed applications. They're sandboxed, single-file, and don't persist data between sessions. Use them for the 70% solution and move to proper development when you need the remaining 30%.

Try Claude-in-Claude. Ask Claude to build an artifact that itself uses the Claude API. The first time you see a tiny app answering questions, generating content, or processing text using AI running inside another AI's output — it clicks.

The Category Change

I had spent years creating static slide decks to explain complex concepts. Hours in PowerPoint, arranging boxes and arrows, writing speaker notes, rehearsing transitions. And then one prompt to Claude produced something that was more engaging, more interactive, and more useful — in ninety seconds.

That's not an efficiency gain. That's a category change.

I still make slide decks. I still write documents. But whenever I can, I reach for an artifact instead. Because the moment you give someone an interactive tool to explore an idea, instead of a static picture to stare at, the quality of the conversation that follows is fundamentally different.