Compound Engineering 3.28 ships cross-model dispatching with Grok 4.7 xHigh default
kieranklaassen · x · 2026-09-22
Every's Compound Engineering coding-agent skill set ships v3.28:
- Cross-model dispatching: skills can now dispatch to Grok 4.7 defaulting to xHigh reasoning effort, covering cross-model code review and doc review
- Configurable reasoning effort for external workers (previously pinned to high)
- OpenCode V2 compat fixes and ce-babysit-pr fixes
The project is a set of skills layered onto your existing coding agent (two commands in Claude Code; also runs on Cursor, Codex, Kimi Code CLI, Devin, and 20+ other hosts). Its core loop is Plan → Work → Review → Compound: learnings are captured after each fix and read back by future plans, so each unit of work makes the next one easier. The catalog spans the core loop (ce-ideate/plan/work/compound), on-demand skills like ce-code-review and ce-debug, Git workflow skills, and an autonomous pipeline.
More from coding & agent
- Teknium merges fix for Hermes Agent Desktop failing to resolve model-provider plugins — Teknium · 2026-09-22
- Grok 4.7 jumps to 46.3% on CursorBench and 64% on EEBench, keeping the same $2/$6 per million token pricing — FinanceYF5 · 2026-09-22
- OpenCode 2.0 Blog Reveals an Agent That Rewrites Itself Live During Sessions — aidenybai · 2026-09-22
- Tests Passing ≠ Ready to Ship: Reddit Debates Whether AI Coding Agents Should Merge Their Own PRs — Fantastic-Sleep-3352 · 2026-09-22
- Scale AI's Alexandr Wang hypes user giving Muse agent access to their bank account — alexandr_wang · 2026-09-22
- arcstone-mcp-sidecar: A Rust Sidecar That Blocks MCP Tool Call Replay Attacks with Atomic File Locks — AdmissibilityScience · 2026-09-22