How Claude Code Built a Working System in One Hour – Lessons from Google Engineers

The rapid rise of AI coding tools

Jaana Dogan, a chief engineer for Gemini at Google, explained that she described a loosely defined problem to Anthropic’s Claude Code and received a working prototype within an hour—something her team had been developing for a year. The task involved a distributed‑agent orchestrator, a system that coordinates multiple AI agents.

Dogan noted that internal Google attempts never reached consensus, whereas Claude Code produced a simplified version that demonstrated the current capabilities of AI‑generated code. She cautioned that the output is not perfect and needs further refinement, but it showcases how far AI‑assisted programming has progressed.

Timeline of AI‑assisted programming milestones

2022 : Systems could generate a single line of code.

2023 : Tools began generating entire code sections.

2024 : Applications could be built across multiple files.

2025 : Full codebases could be created and refactored automatically.

These advances are reshaping software development, shifting developers from writing every line to reviewing and orchestrating AI‑generated output.

Claude Code creator shares workflow tips

Boris Cherny, a developer of Claude Code, recommends letting the model verify its own work, which can improve output quality two‑ to three‑fold. He suggests starting most tasks with a planning phase, iterating with Claude until the plan is solid, then letting the model execute the task in one pass.

For repetitive workflows, Cherny uses slash commands and sub‑agents to automate specific actions such as code simplification or testing. Long‑running tasks are handled by background agents that run multiple Claude instances in parallel, reviewing results after completion. His default model is Opus 4.5.

In code reviews, his team tags Claude directly in pull requests to add documentation. Claude Code also integrates with external tools like Slack, BigQuery for data analysis, and Sentry for error logging.

Google and Anthropic partnership

Google is Anthropic’s largest investor, having contributed roughly $3 billion. In October 2025 the two companies deepened cooperation, with Google agreeing to provide up to one million dedicated TPUs, a deal worth hundreds of billions of dollars.

The collaboration is expected to deliver over 1 GW of AI‑training capacity by 2026, representing the largest hardware investment in the AI industry to date. Anthropic chose Google’s TPUs for their cost‑effectiveness and performance.

What’s next?

Both Google and Anthropic are racing to improve AI coding capabilities. Dogan confirmed that her team is actively enhancing Gemini’s models and infrastructure to stay competitive. The expanded TPU agreement will dramatically increase compute resources, accelerating the next generation of AI‑coding tools.

As AI‑written code becomes commonplace in large tech firms, human developers will increasingly focus on higher‑level supervision, architectural decisions, and quality control rather than hand‑crafting every line.