My coding agents went from 30 min runs to 8 hour overnight runs in 5 months on t

Somewhere between November 2024 and April 2025, a developer running a coding agent workflow noticed something that didn't announce itself dramatically — it just accumulated. Thirty-minute runs became sixty. Sixty became several hours. By April, the same workflow that once finished before a lunch break was running overnight, clocking eight hours on tasks structurally identical to what it had always handled. The same inputs, the same architecture, the same expected outputs. Only the time had changed, and it had changed by a factor of sixteen. The post wasn't a rant. It was a measurement, held up with the quiet unease of someone who has checked their instruments twice and still doesn't like the reading. Five months of documented creep, shared into a forum, generating one recorded signal so far.

If confirmed, here is what this means. The degradation of AI agent performance over time — not through model updates, not through workflow changes, but through some ambient drift in underlying system behavior — would represent a serious accountability gap in how these tools are sold and supported. Developers building production pipelines around agent-based coding assistance are implicitly trusting that baseline performance holds. A sixteen-fold slowdown doesn't just waste compute costs; it breaks the scheduling assumptions that make overnight automation viable in the first place. The second-order effect is more unsettling: if performance can erode this slowly and this silently, the developers most likely to miss it are the ones who integrated the workflow early and stopped watching it closely. Trust, once established, tends to reduce vigilance. That's precisely when slow degradation does its damage. For enterprises evaluating agentic coding tools, this pattern — if it holds up — argues for mandatory performance benchmarking at regular intervals, not just at onboarding.

Watch for corroborating posts from other developers reporting similar timing drift on stable workflows, and for any response from OpenAI acknowledging changes to inference prioritization or context-handling that might explain the pattern.

NewsHive monitors these sources continuously. All signal titles above link to the original reporting.

Intelligence by NewsHive. Need help navigating what this means for your business? Contact GeekyBee →