No topics yet. Start the conversation.
Summary
Count of publicly announced plan-tightening actions per vendor, cumulative.
Description
The canary thesis
Agentic coding sessions burn 10–100× more tokens per user than chat did, so every vendor selling a flat-rate seat is quietly losing money on heavy users. The thesis: if every agentic vendor tightens within a 90-day window, flat-rate is dead. This chart tracks the canary events.
Signal strength: 34 tightening events across 7 vendors (Amazon, Anthropic, Cursor, GitHub Copilot, JetBrains, OpenAI, Windsurf) since the first move in May 2025. The March-April 2026 cluster (Windsurf overage + OpenClaw block) is a second clear tightening window.
Event log
Full timeline of every pricing, metering, and harness-restriction action we track. Tightens = Y means the action transfers cost or risk from vendor to user. Pre-2025 rows are baseline context and don't count toward the chart's cumulative lines.
Source: vendor blogs, changelogs, and docs. Stored in Turso ai-inference-unit.events; refreshed nightly by the ai-inference-unit-refresh Novem job.