Why Everyone Is Suddenly Obsessed With Claude Code

Claude Code looks like the opposite of a modern, polished app. It runs in a terminal, it’s black-and-white, and it can be a pain to install. Yet people are genuinely obsessed with it—many of them not even developers. They’re using it late into the night to build apps, automate workflows, and even book flights.

So what makes this tool so compelling? Under the hood, Claude Code quietly combines powerful AI, clever design psychology, and a deep understanding of how people actually like to work.

Claude Code Feels Like a Strategy Game, Not a Tool

Using Claude Code doesn’t feel like filling out a form or chatting with a bot. It feels more like playing a “just one more turn” strategy game. You give it a prompt, it thinks, takes actions, edits files, runs commands, and then shows you the result. Then you tweak the prompt and go again.

This works because it accidentally taps into some of the strongest principles from game design:

Short, Satisfying Feedback Loops

Great games are built from many small loops of action and reward. You do something, see the result quickly, and decide what to do next. Claude Code works the same way: you describe what you want, it plans, edits files, runs tests, and shows you progress within seconds or minutes.

Each prompt is its own mini “level.” That fast loop keeps you engaged and makes it very easy to say, “I’ll just try one more change.”

Juicy, Playful Feedback

In game design, “juiciness” means adding small, fun details that make interactions feel alive—like screen shakes, sound effects, or little character animations. Claude Code does a low-key version of this.

Claude’s tone is friendly and conversational, not stiff or robotic. And when Claude Code is working, it shows playful status messages and progress text. Thanks to the leaked source code, we know there are 187 of these little phrases baked in. They’re not necessary for functionality, but they give the tool personality and make it feel more fun to use.

Visible Progress Everywhere

Games constantly show you progress bars, XP meters, and maps filling in. Claude Code does something similar with your project itself. You start with an empty folder, and as it works, files appear, directories grow, and features come to life in real time.

There’s no natural stopping point—it’s like an open-world or strategy game. There’s always one more feature to add, one more bug to fix, one more idea to try. That sense of ongoing progress is a big part of why people lose track of time while using it.

A Bit of Unpredictable “Luck”

Claude is powerful, but not perfect. Sometimes it produces something brilliant, sometimes just okay, and sometimes it gets close but not quite right. That unpredictability actually makes it more addictive.

When you’re almost there, it’s very tempting to think, “I’m so close—just one more prompt.” It’s the same psychology that makes slot machines and loot boxes so compelling: you never know when the next attempt will be the perfect one.

The “Labor Illusion”: Watching It Work Is Rewarding

Many consumer AI tools hide their work behind a simple spinner. Claude Code does the opposite. You see it reading files, running commands, hitting errors, fixing them, and trying again.

Researchers call this the “labor illusion”: when you see visible effort, you value the result more. Watching Claude Code grind through tasks makes the outcome feel more impressive and satisfying, even if you didn’t do the work yourself.

Why a Hard-to-Use Terminal App Feels So Simple

On paper, Claude Code should be intimidating. It runs in the terminal, exposes raw commands, and doesn’t hide complexity behind a pretty UI. Yet many people who aren’t developers find it easier and more powerful than polished “AI for everyone” apps.

The key idea here is Tesler’s Law: complexity can’t be destroyed, only moved. A truly simple-feeling product doesn’t remove complexity; it just hides it in the right place.

Perceived Simplicity vs. Real Complexity

Think about copy-paste. For you, it’s just Ctrl+C and Ctrl+V. Behind the scenes, your system is juggling clipboard formats, images, rich text, and even syncing between devices. It feels simple because all the complexity is hidden under the hood.

Many modern AI products get this wrong. They’re visually simple but not very powerful. They can only do a narrow set of things, so they feel limiting after the first few uses.

Claude Code does the opposite. It’s genuinely powerful and flexible, but it pushes as much complexity as possible onto the AI itself. When something breaks, it tries to fix it. When it needs documentation, it goes and finds it. When a tool doesn’t work, it switches strategies.

From the user’s perspective, you just describe what you want. Claude Code deals with APIs, servers, errors, and obscure configs in the background. That’s perceived simplicity at work.

Self-Healing Workflows

One of the most impressive aspects is how Claude Code can debug its own environment. For example, if it tries to use an MCP server and fails, it can:

• Notice the failure
• Search for alternative approaches (like using an API directly)
• Read online docs
• Reconfigure its own setup to make things work

You don’t have to know what an MCP server is or how to fix it. You just say what you’re trying to do, and if something goes wrong, you ask Claude Code to “unstuck” itself. The system is complex, but your experience stays simple.

If you want to go deeper into how this works in practice—models, MCP, skills, and more—check out this in‑depth Claude Code guide.

Why Non‑Developers Are Using a Coding Tool for Everyday Tasks

One of the strangest things about Claude Code’s rise is who’s using it. Yes, developers love it. But a huge chunk of users are curious non‑technical people who follow tech and just want to get things done.

They’re using Claude Code to:

• Book airline tickets with complex constraints
• File taxes
• Monitor tomato plants
• Search through music archives
• Clean up and organize huge folders of files

Why would they choose a terminal-based coding tool over polished consumer apps?

Everything Is “Code” Under the Hood

Most digital tasks you do are, at their core, just code with a UI on top:

• Moving folders is a file system command
• Renaming 500 photos is a script
• Analyzing a spreadsheet is code running on tabular data
• Searching flights is a set of API calls with filters

Traditional apps lock you into whatever options their interface designers imagined. For example, a flight search site might let you filter by dates and stops, but not “one-day layover in Singapore, then connect to Taiwan, with options to continue to the US.”

Claude Code doesn’t care about the UI. It can call APIs directly, read data from files, write scripts, and stitch everything together. You describe the outcome, and it figures out which tools and code to use.

This doesn’t teach people to code in the traditional sense. It removes the barrier between them and the code that was always there, hidden behind interfaces.

What the Source Code Leak Revealed

The leaked source code for Claude Code gives a rare look into how it actually works behind the scenes—and why it feels so capable.

A Toolbox of Actions, Chosen Automatically

When you ask Claude Code to do something, it doesn’t just generate text. It:

• Scans your project and files to understand the current state
• Creates a plan of action
• Chooses from 30+ different actions (read files, write code, run commands, search the web, etc.)
• Executes those actions step by step

If something fails, it reads the error, adjusts its approach, and tries again. This creates an “infinite loop” of planning, acting, and correcting until the job is done—or it hits a real dead end.

There are also clever tricks in the implementation: multiple strategies for compressing context, fake tools added to confuse competitors, and detailed logging of user frustration (including when you swear at it) to measure how well it’s doing.

Smart, but Also Surprisingly “Dumb”

Despite the sophistication, a lot of the system is surprisingly blunt. The prompts are often just begging the model to follow instructions precisely. The magic comes less from a single genius trick and more from carefully combining:

• A capable language model
• A rich set of tools and actions
• A tight action–feedback loop
• A design that exposes progress instead of hiding it

That combination is what makes Claude Code feel like a tireless junior engineer—or a powerful AI agent—rather than just another chat box.

If you’re new to Claude and want to go from first prompt to using Claude Code effectively, you might find this beginner-friendly Claude tutorial helpful.

The Hidden Cost of an Addictive AI Coding Tool

There is a downside to building a tool that works this well: it’s expensive to run. The highest Claude Code tier is around $200 per month, and heavy users can easily keep multiple instances running in parallel all day.

In one extreme case, a single user consumed an estimated $27,000 worth of compute on a $200 subscription. That means the product is so compelling that people are pushing it to its limits—and the company is likely losing money on its most dedicated users.

Claude Code shows what happens when you combine strong AI models with smart product thinking: you get something that feels less like a chatbot and more like an endlessly helpful collaborator. It’s rough around the edges, lives in the terminal, and makes mistakes—but that might be exactly why people can’t stop using it.