AI Made Coding Cheaper, So I Work Harder Than Ever

The author, now VP of engineering at Bold.org, expected AI agents to reduce his workload. Instead, running multiple agents in parallel has expanded it dramatically. Each agent periodically requires input—to review code, approve plans, or interpret results—forcing constant context switching between different problems. This mental fatigue differs from his previous pattern of focused work on two or three tasks per week.

GitHub contribution history quantifies the shift. During his peak full-stack engineering years (2020), he recorded 2,711 contributions. As a Principal Engineer from 2022-2025, this fell to an average of 114 per month. His VP role beginning August 2025 initially continued this lower pace. But 2026 reversed the trend: by October 27, he had accumulated 2,102 contributions—nearly matching his 2020 high—with August alone at 427, his busiest month ever. His median weekday contributions rose from 3-4 to 7.

The expansion reflects how inexpensive AI made implementation. Rather than freeing time, cheap coding enabled him to pursue projects previously considered not worth starting. This triggered recurring cycles of bursts followed by burnout. While most 2026 quiet periods showed elevated activity—March had 203 contributions versus a 114-per-month baseline—June’s 77 contributions fell below that average, showing the pattern is not uniform across all months.