95-100% of Coinbase's Code Written or Assisted by AI, Engineers Manage 5-10 AI Agents
Rob Witoff, head of the Coinbase platform, stated that currently 95% to 100% of the company's code is written or assisted by large language models, a significant increase from about 40% in February of this year, with nearly all employees using AI daily.
Core cryptography still relies on manual review, while internal prototype development is nearly 100% automated, allowing 2-3 person teams to accomplish the work previously done by over 10 people.
Most engineers manage 5 to 10 AI Agents simultaneously, with the overall workload of AI Agents equivalent to that of 1,200 employees, expected to rise to the equivalent of 100,000 employees by 2030. This year, layoffs of about 700 people accounted for 14% of the workforce.
In terms of market mechanisms, AI enhances efficiency, driving cost optimization and productivity leaps, shifting tech company valuation logic from human scale to AI leverage, with increased demand for related tools and infrastructure.
Source: Public Information
ABAB AI Insight
Coinbase is actively embracing AI as a crypto platform, similar to its early technological iterations, which enhanced the efficiency of trading and compliance modules through automation tools.
On the capital path, the company's engineering resources are shifting towards AI Agent management and code generation, with layoffs occurring alongside a significant increase in workload, strategically maintaining high output with smaller, more experienced teams to accelerate product iteration.
Similar to the AI transformation paths of tech companies, Coinbase is currently in the accelerated phase of code generation and Agent automation.
This represents a leap in productivity; the surge in AI code writing and Agent workload at Coinbase is due to LLMs lowering coding barriers and amplifying the leverage of senior engineers, thereby achieving workforce streamlining and efficiency multiplication. By 2030, the equivalent of 100,000 employees is expected to reshape the entire industry's human capital structure.
ABAB News · Cognitive Law
- Doubling AI code proportion leads to exponential productivity increase.
- Small teams + multiple Agents outperform large-scale human labor.
- The equivalent of 100,000 employees by 2030 marks the full explosion of AI leverage.