Ship It or Skip It #001: Meta enters the coding-agent race

Week of Oct 8–15, 2026 · Tested picks for vibe coders and agent builders

Sample draft. Verdicts are provisional: they come from the release notes, and our hands-on tests haven't run yet. Swap in test results before sending.

TL;DR

⭐ FEATURED (Sponsored)

[Your tool here] — 50 words, one link, shown to every reader. Founding rate: $49/issue. Reply to book.

(Slot reserved for a paid or partner placement; label "Sponsored" stays.)

🚢 SHIP

Codex CLI 0.162.0 (Oct 8)

What changed: managed Git worktree tools, pinned tasks in the agent Command Center (p), /copy for transcript blocks, clickable URLs in approval prompts, apply_patch now keeps CRLF line endings, Linux/Windows sandbox fixes.

Why it matters: worktrees let you run parallel agents without them stepping on each other's files.

Do this: update, then turn on the worktrees feature and give each agent task its own worktree.

Source: https://github.com/openai/codex/releases/tag/rust-v0.162.0

Cursor CLI v2026.10.08 (Oct 8)

What changed: Auto-review now asks before risky recursive/wildcard deletes. Claude Opus 5.5 and Sonnet 5.5 are pickable on Bedrock bring-your-own-key. /skills flags skills that break the spec limits. Agents in ACP mode now wake when a watched background shell prints a match. Big batch of tool and MCP fixes.

Why it matters: fewer "the agent deleted my folder" moments, and long sessions recover from hitting the context limit.

Do this: agent update, then run /skills and fix any yellow warnings.

Source: https://cursor.com/docs/cli/changelog

🧪 TRY

Meta Muse Wonder / Muse Wonder Full (Oct 12)

What it is: Meta's frontier coding and agent models. Muse Wonder costs $4 input / $20 output per million tokens with 1M context on the Meta Model API and Muse Code. Muse Wonder Full is only available on Muse Code's Power and new Maximum Usage plans.

Claims: Meta says Full tops the DeepSWE and SWE-Marathon leaderboards in third-party runs. These are vendor-reported; we haven't verified them.

Who it's for: long-running agent tasks and multi-file refactors. Pricier per token than you might expect, so test on a real task before switching.

Source: https://dev.meta.ai/resources/blog/build-with-muse-wonder

JetBrains Mellum2.1 (Oct 12)

What it is: an open 12B mixture-of-experts coding model (2.5B active parameters, Apache 2.0) trained for the explore, edit and check loop. Runs via GGUF on llama.cpp, Ollama and LM Studio.

Who it's for: private repos and laptop-local agents. Speed claims come from JetBrains, so run it on one of your own issues first.

Source: https://madewithmodels.com/news/2026-10-12/jetbrains-mellum21-coding-agents/ (secondary coverage; link JetBrains' own post before sending)

Cursor Origin App Marketplace (Oct 13)

What it is: Origin now connects to 20+ apps (Vercel, Netlify, Cloudflare, Sentry, Linear, CodeRabbit, Semgrep, Socket and more) and accepts public app submissions.

Who it's for: teams wiring deploy, CI and review bots into their agent workflow.

Source: https://cursor.com/changelog

⏭️ SKIP

👀 Watchlist

🗂️ Directory spotlight

New this week in the Ship It or Skip It directory → [link]. Building a tool? List it free, or fast-track it for $39.

Forward this to a friend who ships. · Sponsor an issue · Unsubscribe