AI Already Writes 90% of Code: What It Really Changes for Your Tech Team

_{IN ONE SENTENCE}

At Anthropic, 90% of code lines are now AI-generated. Dario Amodei details what this concretely means for technical teams and why the developer role is transforming radically.

_{THE OBSERVATION}

It's done. 90% of lines written by the model; an unprecedented productivity leap in software development. But an essential nuance must be added: it's not simply about coding faster. Models now perform end-to-end development tasks including compilation, testing, and deployment.

Even more surprising: the trend extends beyond pure code. Tasks like documentation, environment configuration, and technical memos, all performed by models. 100% of development tasks now integrate AI to some degree.

_{WHAT YOU NEED TO UNDERSTAND}

Coding is the first domain transformed

Amodei explains why code was the first domain where AI reached near-professional level. Code offers objective verification: code compiles or doesn't, tests pass or fail. This rapid feedback loop allows models to improve much faster than in domains where quality is subjective.

What's coming for all professions

The pattern observed in coding will repeat across all domains of intellectual work. Writing, financial analysis, consulting, legal; everything will be affected following the same pattern, with a slight time lag. The question for each profession is no longer if, but when.

The developer's new role

Developers aren't jobless; new tasks in supervision, architectural design, and prompt engineering are emerging. But the profession is fundamentally transforming. What matters now isn't typing speed or syntactic mastery. It's the ability to design, supervise, and correct systems. The developer becomes an orchestrator of AI systems, not a producer of code lines.

Metrics must change

If AI generates 90% of code, measuring productivity in lines of code is meaningless. What matters is the outcome: feature throughput delivered, final product quality, speed of problem resolution. Companies that don't change their metrics will optimize for the wrong thing.

_{WHAT THIS CHANGES FOR YOU}

• Transform your development processes now. 90% AI generation is not a projection: it's the current reality at technology leaders.
• Train your developers in supervision and architecture skills rather than pure coding. The profession is evolving toward orchestration.
• Change your technical productivity metrics. Measure features delivered and product quality, not lines of code.
• Prepare for extension to other professions. What's happening in coding today will happen in all domains of intellectual work tomorrow.