In a surprising trend, developers are refusing to work without AI coding tools, but research suggests this reliance may be costing them more than it's worth.
What Happened
Researchers at METR, an acclaimed AI research institute, attempted to repeat a groundbreaking study on AI coding productivity in February 2026. However, they were unable to do so because developers refused to participate even for a limited number of tasks without the aid of AI.
This refusal is a stark contrast to the original 2025 study, which found that developers reported increased productivity when using AI tools. However, the data revealed that AI actually slowed them down due to the extra time spent finding and fixing errors, steering the AI, and waiting for tasks to complete.
Unable to replicate the experiment, METR published a survey in May instead, allowing tech employees to self-report their AI productivity gains. Not surprisingly, developers believed that AI doubled their value to their organizations. However, recent evidence from multiple sources suggests that this perception is wrong.
Background and Context
The trend of relying heavily on AI coding tools has been dubbed "tokenmaxxing," where the number of tokens used as a proxy for productivity is prioritized over actual output quality. This approach has been adopted by several high-profile companies, including Amazon and Uber.
Amazon recently shut down its internal token-tracking leaderboard, Kirorank, after employees were found to be gaming the system by using AI agents excessively and running up costs. Similarly, Uber blew through its 2026 AI budget in just four months, with executives admitting no measurable gains in output or productivity.
The code quality problem is a deeper issue that has been highlighted by programmer and author James Shore in a viral blog post. He argued that faster code generation without reduced maintenance costs is a trap, where developers are essentially signing up for permanent overtime.
Why It Matters to the Industry
The reliance on AI coding tools and the focus on tokenmaxxing have significant implications for the adult industry. With the increasing demand for high-quality content and the need for efficient production processes, companies may be tempted to adopt similar approaches.
However, this trend raises concerns about code quality, maintenance costs, and the potential for long-term maintenance debt. As AI-generated code creates more problems than human-written code, developers may find themselves struggling to keep up with the demands of their jobs.
The adult industry is particularly vulnerable to these issues due to its high demand for content and the need for efficient production processes. Companies that adopt tokenmaxxing approaches without considering the potential consequences may find themselves facing significant challenges in maintaining quality and reducing costs.
What Comes Next
Experts recommend treating AI output like junior developer work, requiring human oversight and review to ensure quality. Advanced AI coding agents like Devin, designed to fix AI bugs on the fly, are not a solution but rather a temporary fix that requires constant monitoring.
The industry must adopt a more nuanced approach to AI adoption, prioritizing code quality over tokenmaxxing and focusing on long-term maintenance costs. By doing so, companies can ensure that their reliance on AI coding tools does not come back to haunt them in the future.
Key Facts
- Developers are refusing to work without AI coding tools, even for a limited number of tasks.
- The original 2025 study found that AI actually slowed down developers due to extra time spent finding and fixing errors.
- Amazon shut down its internal token-tracking leaderboard, Kirorank, after employees were gaming the system.
- Uber blew through its 2026 AI budget in just four months with no measurable gains in output or productivity.
- AI-generated code creates more problems than human-written code, requiring extra time and resources for maintenance.
- Experts recommend treating AI output like junior developer work, requiring human oversight and review to ensure quality.
