Share with your CTO
Anthropic’s Claude Code origin story traces how a VS Code extension that sat dormant for most of 2022 became the fastest enterprise software ramp on record: $1 billion in annualized revenue within six months of its May 2025 general availability launch, $2.5 billion by February 2026, and an estimated 4% of all public GitHub commits authored by the tool daily. Eight of the ten Fortune 10 companies now use Claude. In a February 2026 survey of 906 developers, Claude Code led satisfaction at 46% compared to 9% for GitHub Copilot, despite Copilot’s four-year head start and Microsoft’s full enterprise distribution machine behind it.
What this means for your business
The 46-versus-9 satisfaction gap isn’t a preference score. It reflects a categorical difference in what the tools actually do. GitHub Copilot suggests the next line of code as a developer types. Claude Code receives a task description, reads the relevant files, writes across multiple files, runs tests, executes shell commands, and iterates until the task is complete. The practical consequence: a developer using Claude Code isn’t slightly faster at writing code. They’re doing a different job entirely.
Boris Cherny’s development progression tells the real story: Claude Code wrote 10% of his code in February 2025, 40% by May, and 100% by winter. He now ships up to 88 commits in a day without writing a line by hand. The compounding dynamic here is what CTOs need to price into hiring and tooling decisions right now. Teams that built Claude Code workflows in mid-2025 have already accumulated months of institutional practice that late adopters are starting from zero on. That gap widens with every model release.
The security record complicates the enthusiasm, and it should. A prompt injection vulnerability in the Claude Code GitHub Action, documented by Microsoft Security researchers in June 2026, can expose CI/CD pipeline secrets when the agent processes untrusted GitHub issue content. Anthropic patched it, but the vulnerability class doesn’t disappear with a patch. Any pipeline where Claude Code touches both untrusted public input and privileged credentials is a distinct risk category from an engineer using the tool interactively. The signal worth watching: whether Anthropic’s security disclosure cadence keeps pace with the release cadence as the tool moves deeper into automated enterprise workflows.
Concept deep-dive: Agentic loop
An agentic loop is the architectural pattern that separates tools like Claude Code from autocomplete. The model receives a goal, evaluates current state (reads files, checks test output), picks an action (write code, run a shell command, call an API), observes the result, and repeats until the task is done or it gets stuck. Think of it as the difference between asking a contractor to suggest paint colors versus handing them a key and a punch list. The business implication: agentic loops can operate unsupervised, which is exactly why the authorization model around them matters so much.
Based on reporting from Claude Code Origin Story: How a Forgotten CLI Became the Fastest Enterprise Software Ever, originally published 2026-07-14 15:07:00.
