Tenex scales output-based pay for 75 AI engineers using Git-tracked MetaHarness

To coordinate engineering with AI agents, Tenex developed MetaHarness, a CLI and repository structure where architectural decisions, plans, and acceptance criteria are checked into Git alongside source code. Hess explained that having agents document context directly in the repository ensures that subsequent automated tasks retain the reasoning behind previous decisions. According to Hess, engineer output roughly doubled over the past year, driven by long-running task capabilities in frontier models like Opus 5.5, which has become the firm's default tool alongside AMP and Codex.

Lieberman observed that enterprise adoption continues to face practical bottlenecks. While roughly 90% of the approximately 100 enterprises Tenex communicates with have deployed general-purpose models, only about 60% provide structured training, and roughly 50% have introduced coding agents to engineering teams. Lieberman also voiced skepticism toward specialized vertical AI software products, arguing that generalized models risk absorbing point-solution features over time.