Fetching from the wire…
Tools2026-08-13 · source-backed
Cursor shipped two modes, Auto Intelligence and Auto Balance, routing each request based on production traffic feedback rather than static rules. This is the heterogeneous-routing pattern practitioners have hand-built for a year, expensive model for reasoning, cheap for mechanical subtasks, moved into the editor as a managed default. The tradeoff you're accepting: you no longer know which model wrote a given diff.
Each link below shares sources, entities, or timing with this story.
Cursor's August 6 post details a two-stage router: "Compass" assigns each turn a 0–1 complexity score by predicting whether the user will be satisfied, then a taxonomy across domains (backend, frontend, database), tasks (bug fixes, commands, tests), and modifiers (visual chang...
Three things happened almost simultaneously. Cursor 3.3 shipped "Build in Parallel," which identifies independent parts of your plan and runs them concurrently using async subagents. Windsurf integrated Devin Local with cloud handoff and multi-model support. And Claude Code's...
Within five days of each other, both Claude Code (v2.1.158, May 31) and Cursor (3.6, May 29) shipped remarkably similar architectures for autonomous agent execution. Both use a classifier subagent that reviews each pending action against conversation context and decides: allow...
The coding agent wars just entered a new phase. Cursor isn't just an IDE anymore. It's a model company. Cursor released Composer 2.5 on May 18 with a custom agentic coding model trained using 25x more synthetic tasks than Composer 2 and a novel "targeted textual feedback" appr...
Raw coding capability is no longer a moat. Cursor just proved it with numbers that are hard to argue with. Cursor released Composer 2.5 on May 18, built on Moonshot AI's open-weight Kimi K2.5, a 1-trillion-parameter mixture-of-experts model that activates just 32 billion param...
xAI launched Grok 4.5 and Grok Build on July 8, trained partly on Cursor developer-session data. The numbers are loud: 83.3% on Terminal-Bench 2.1, 64.7% on SWE-Bench Pro, priced at $2/$6 per million tokens. On a single coding task that works out to roughly $2.49 versus $11.80...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.