AI Now Writes Half of All Linear Issues  -  and Teams Still Ship

Linear's aggregated data across all paid workspaces shows AI now authors just under half of all issues created on the platform, up from fewer than one in a thousand two years ago, yet total product development time increased. Teams using coding agents tripled their weekly pull requests from 21 to 65, while non-agent teams rose from 8 to 10, but cycle time grew, indicating agents added work rather than removing it. Linear's report, published August 20, 2026, contradicts the AI productivity narrative, suggesting coordination overhead now falls on humans.

AI Now Writes Half of All Linear Issues - and Teams Still Ship Linear published aggregated data across its paid workspaces. AI authors just under half of everything created, up from fewer than one in a thousand two… The most important number in AI this week isn't a benchmark /glossary/benchmark score. It's a product metric, and it says something the industry has spent two years trying not to hear. Linear published aggregated data across every paid workspace on its platform, comparing June 2025 against June 2026. The headline: AI now writes just under half of everything created on the platform. Two years ago, it was fewer than one issue in a thousand. And then the other number. Total product development time went up anyway. The Paradox, Measured Let that sink in. Teams running coding agents tripled their weekly pull requests, from 21 to 65. Teams not using agents went from 8 to 10. The people who adopted the tools are doing three times the output per week. And their total development time rose, not fell. So did time spent creating and triaging issues across nearly every function. Founders swung hardest: up 17 minutes on issue creation and 26 on commenting. Here's the thing nobody wants to admit out loud. A team tripling its pull request throughput while its cycle time grows is not a team going faster. It's a team doing more things. The agents added work rather than removing it. Where the Time Went Think about what actually happens when an agent opens three times the pull requests. Somebody has to review all of them. Somebody has to write the context the agent acts on. Somebody has to triage the issues the agent filed, many of which are noise. That coordination overhead landed on the exact same engineers who were supposed to be freed up. Linear frames it bluntly: more work needs more coordination, and the coordination is now the context agents run on. You didn't remove the bottleneck. You moved it, and it landed on the humans. Sound familiar? It should. This is the same story every productivity-tool boom has told. The spreadsheet didn't kill the analyst's job, it gave the analyst three times the spreadsheets. Why This Is Different From a Bad Benchmark Most AI-industry numbers are vendor figures with a stake in the outcome. Linear sells the tools these teams use, sure. But the data doesn't flatter the AI narrative. It undercuts it. A company shipping agents would love to report that its users got faster, and instead it published evidence they got slower. That's what makes this worth taking seriously. When the raw telemetry contradicts the marketing, the telemetry is probably the honest part. The Counter-Example That Hasn't Shown Up The real question now is whether anyone publishes the opposite case. A team that tripled throughput and cut cycle time would be front-page news. It hasn't surfaced. Its absence is doing a lot of quiet work in this debate. My read: the tools work. The workflow doesn't. Agents are force multipliers, and a force multiplier pointed at the wrong thing multiplies the waste. Until teams redesign review, triage, and context-setting around the agent rather than bolting the agent onto the old process, the paradox holds. You'll ship more code and feel slower doing it. Sources: Linear engineering productivity report, August 20, 2026; AI Tools Recap daily briefing, August 21, 2026. Get AI news in your inbox Daily digest of what matters in AI.