The Production Run

A software-development platform proposes a public, measurable “factory” model for scaling AI-agent work, predicting that by August 2028 more than 20% of developers on its leaderboard will reach a top tier defined by running at least 10 concurrent agent sessions, merging 10 agent-generated pull requests weekly, remaining active most days, and keeping costs below $3.50 per merged PR. Performance is assessed across five non-compensating dimensions—parallelism, token depth per session, merged output, sustained activity, and cost efficiency—so a developer’s tier is determined by their weakest metric rather than an average. The framework describes four tiers, progressing from closely supervising one agent to managing durable, autonomous queues of agents whose work can finish overnight, and projects a shift from nearly all users at the entry tier in 2026 to a median “Plant Manager” level by mid-2028. It acknowledges measurement limitations, particularly incomplete GitHub PR tracking, and commits to revising the instrument publicly without lowering tier thresholds. Its first 30-day production run, beginning in September 2026, challenges all participating developers to reach the second-tier “Operator” standard by using two parallel agents, assigning substantial tasks, merging roughly one agent PR per week, staying active for 10 days, and meeting a cost-efficiency target.

No tracked trend matches for this post yet.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.