In 2026, some software development teams began using token consumption as a key performance indicator (KPI) to measure developer productivity. This practice, dubbed 'tokenmaxxing,' involves tracking the number of AI tokens spent daily. Critics, including DORA in their June 2026 report, classify this as a vanity metric that is easily manipulated and disconnected from actual project outcomes.
The industry is also experimenting with Spec-Driven Development (SDD). While KDDI reported a 3.1-fold increase in deployment frequency—from 4.4 to 13.8 times per week—using this approach, experts warn that SDD can lead to technical debt if the team lacks maturity. Without proper processes, specifications may be created merely to satisfy reporting requirements rather than to improve the product.
Other metrics, such as AI adoption rate and lines of AI-generated code, are also criticized for failing to reflect quality. LinearB data shows that while top teams use AI in 54% of pull requests, less than 5% of those PRs come from autonomous agents. These indicators often mask a lack of real impact, as developers may simply rely on autocomplete tools without achieving meaningful results.
Experts suggest shifting focus toward metrics that measure actual value, such as 'cost per verified outcome'—the cost of a feature that reaches production without issues. Research by AMCIS in 2026 supports a portfolio-based evaluation, which considers speed, quality, and impact collectively. The consensus is that measuring tool usage rather than user-facing results leads to imitation of productivity.
