Claude Code Explained: Why Tech Leaders Call It 'a Religion'

At the biggest AI conference of the year last week, a CEO described a product as “a religion.” He wasn’t being sarcastic.

Arvind Jain runs Glean, a $7 billion enterprise AI company. At HumanX in San Francisco on April 8, 2026, he said Claude Code had inspired what he called “Claude Mania” — a wave of demand so intense that business leaders are being pressured to deploy it even before they understand what it does. His exact words to CNBC: “It has become a religion. That’s the level of that mania.”

If you haven’t heard of Claude Code, you’re not alone. Most non-developers haven’t. But 4% of every public commit on GitHub is now written by it. It’s generating $2.5 billion in annualized revenue. Three out of four AI startups use it.

Here’s what Claude Code actually is, why it’s suddenly everywhere, and — more importantly — whether it matters for your work.

What Is Claude Code?

The simplest explanation: Claude Code is an AI coding assistant that lives in your terminal.

A terminal is the text-based command window developers use — the black screen with blinking cursor you’ve seen in every hacker movie. Instead of clicking buttons, developers type commands. For decades, that terminal has been a solo space: you type, computer responds.

Claude Code changes that. When it’s running, you have a teammate in your terminal. Type a question in plain English — “refactor this function” or “find the bug in the payment flow” — and Claude reads your code, thinks, and writes the solution. It can edit files, run tests, check its own work, and fix what breaks.

This is different from tools like GitHub Copilot, which complete the line you’re typing. Copilot is like autocomplete on steroids. Claude Code is like having a senior engineer pair-programming with you — one who has read your entire codebase in seconds.

Why Is This Such a Big Deal?

Because writing software is hard, and AI that writes software well changes economics.

Here’s what changed between early 2025 and today:

Metric	Early 2025	April 2026	Change
“claude code” monthly searches	~49,500	~1,000,000	20x
Claude Code ARR	Launch (May 2025)	$2.5B+	From 0 to $2.5B in 11 months
% of GitHub commits from Claude Code	~0%	~4%	Projected 20%+ by end of 2026
Startups using Claude Code	Rare	~75%	Near ubiquity
Anthropic valuation	~$60B	$380B	Tripled in a year

Those numbers aren’t marketing. The GitHub commit percentage comes from public git history analysis. The ARR comes from Anthropic’s internal disclosure. The search volume is from Google Ads data.

The short version: Claude Code went from a research project to $2.5B ARR faster than any software product in history. It took GitHub 5 years to hit $1B. It took Claude Code 6 months.

What Happened at HumanX

HumanX is the San Francisco AI industry conference that runs every spring. The 2026 edition, April 6-9 at Moscone Center, brought together 6,500 executives, founders, and investors.

Three reporters covered what they heard. The headlines tell the story:

TechCrunch (April 12): “At the HumanX conference, everyone was talking about Claude.”
CNBC (April 11): “Vibe check from inside AI industry’s main events: ‘Claude mania.’”
Bloomberg (April 9): “Anthropic Dominates the Discussion at the HumanX AI Conference.”

The quotes they captured were vivid. Vendors said they “use Claude a lot” while “ChatGPT and OpenAI had gone downhill.” Executives said deploying Claude Code wasn’t optional anymore — it was a competitive necessity. Enterprise buyers said pressure from their own engineers forced them to evaluate it.

This is a vibe shift. In 2024, every AI conversation started with ChatGPT. In April 2026, the first question is about Claude.

Why Developers Are Obsessed

Three things, in order:

1. It’s actually good. Claude Sonnet 4.6 scores 79.6% on SWE-bench Verified, a standardized benchmark that measures real-world coding ability. That puts it in a narrow lead over GPT-5.4 for most daily coding tasks. More importantly, it hallucinates less — it’s less likely to invent APIs that don’t exist or produce code that compiles but doesn’t work.

2. The workflow is new. Previous AI coding tools forced you to copy-paste code into a chat window, then paste the response back into your editor. Claude Code eliminates that friction. It sits in your terminal where the code already lives. It can edit files directly, run tests, see error messages, and iterate.

3. It has skills, hooks, and MCP. Power users extend Claude Code with “skills” (reusable prompts and workflows), “hooks” (automation triggers on file changes), and MCP connectors (integrations with databases, APIs, documentation). This turns Claude Code from an assistant into a platform — your personal coding environment, customized to your team’s conventions.

The enterprise pattern: companies buy Claude Code for one or two developers, those developers build workflows and skills, the rest of the team watches those developers ship 2-3x faster, and within months the whole company is on Claude Code.

Is This Just Hype?

Partly. But the data is real.

Evidence it’s hype:

Early adopter enthusiasm always exceeds mainstream usage
“70% of developers prefer it” surveys skew toward early adopters on Twitter
The $2.5B ARR is real, but it’s concentrated among tech companies — not representative of the broader economy
Rate limits have been tightening, causing backlash (Anthropic officially admitted the issue on March 31)

Evidence it’s real:

4% of GitHub commits is not hype. That’s a fact about actual code being shipped.
$2.5B ARR doubling in 3 months is not a vanity metric. Someone is paying.
Anthropic jumping from $60B to $380B valuation in a year is investor conviction backed by revenue growth.
The HumanX reaction came from founders and buyers, not Twitter pundits.

Verdict: the hype is real in the sense that the product is legitimately better. The hype is hype in the sense that most of the world doesn’t need an agentic coding CLI, and won’t.

What It Can’t Do

It can’t write software without you. Claude Code is very good, but it makes mistakes. Production deployment without human review is a bad idea.

It needs a terminal. If you’ve never opened one, there’s a learning curve. You need to know basic commands: cd (change directory), ls (list files), git basics. Most developers already do. Non-developers will find the onboarding rough.

It’s not free. Free tier doesn’t include Claude Code. Pro starts at $20/month, and heavy users pay $100-200/month for Max plans.

Rate limits are a real constraint. Even on the $200/month Max 20x plan, heavy users have reported hitting limits. Anthropic acknowledged the problem in March 2026.

It’s not for everyone. If your job doesn’t involve writing code, Claude Code isn’t your tool. Use Claude Cowork for desktop work or ChatGPT for general tasks.

How Non-Developers Can Benefit Without Using It Directly

Here’s what’s interesting: the “Claude Mania” narrative matters even if you don’t code.

If you work at a software company: Your developers will want Claude Code. Prepare for budget requests of $100-$200/month per developer. If you’re a manager, the productivity upside (2-3x shipping speed, per reported usage) usually justifies the cost quickly.

If you hire developers: Knowing Claude Code has become table stakes for senior developer candidates. Resumes are starting to list it as a core skill.

If you’re a business owner evaluating AI tools for your team: The Anthropic ecosystem (Claude web, Claude Code, Cowork, Dispatch, Managed Agents) is broader than “a chatbot.” If you’re planning AI adoption across engineering, operations, and knowledge work, Anthropic increasingly offers a full-stack answer.

If you’re a knowledge worker curious about AI agents: Start with Claude Cowork — it’s the non-developer version of what Claude Code does for engineers. Same underlying AI, different interface.

What This Means for You

If you’re a developer not using Claude Code yet: Try it. A month of Pro at $20 is a low-risk experiment. Install with npm install -g @anthropic-ai/claude-code, run claude auth, and point it at a real project. See what happens. Our Claude Code Mastery course walks through the setup and advanced workflows.

If you’re already using Claude Code: Consider whether your tier matches your usage. Power users are reporting value at $200/month Max 20x. Moderate users often find Pro at $20 plenty. Don’t overpay for capacity you’re not using.

If you manage a software team: Budget for Claude Team Premium at $100/seat (annual). Then ask your most senior developer to build 3-5 custom skills for your team’s conventions. That’s what turns Claude Code from “AI tool” into “your team’s superpower.”

If you’re a non-developer: You don’t need Claude Code. But you might benefit from understanding that the AI productivity revolution is happening fastest in software engineering right now. The patterns being tested there — agentic workflows, custom skills, MCP integrations — are coming to every knowledge worker tool within 12 months.

The Bottom Line

Claude Code is the fastest-growing software product in history. It’s also a $20-$200/month tool that only helps you if you write code for a living.

Those two things can both be true. The “religion” isn’t about Claude Code being the most important product in the world — it’s about Claude Code being genuinely, measurably better at a narrow but economically massive task: writing software.

If you’re a developer, the honest answer is that you should at least try it this month. The worst case is you spend $20 and learn something. The best case is you ship code meaningfully faster.

If you’re not a developer, file this under “important context for understanding how fast AI is changing.” The same pattern — AI that doesn’t just chat but actually does the work — is coming to your industry next.

Ready to try it? Our Claude Code Mastery course takes you from installation to advanced agentic workflows. For the bigger picture of how AI agents fit together, check out AI Agents Deep Dive.

Sources:

Claude Code Explained: Why Tech Leaders Call It 'a Religion'

Table of Contents