LinearB's 2026 Engineering Benchmarks reveal that AI-assisted pull requests merge at less than half the rate of human-authored code, despite 88%+ developer AI adoption. The core issue: AI accelerates code generation but exposes bottlenecks in review processes and unclear requirements. The newsletter also covers MCP's evolving relevance, context anchoring to combat AI memory loss in long sessions, how AI amplifies existing workflow bottlenecks rather than solving them, and emerging workplace etiquette issues around sharing unverified AI output.
Table of contents
1. The rise and fall and rise again of MCP2. Displacing context compaction3. AI will amplify your bottlenecks4. Copilot here, Claude there... now what?5. The resurgence of the small web6. The new workplace etiquette problemSort: