📜 Anthropic just launched a limited-time promo: using Claude's Artifacts feature (documents, presentations, designs) burns half the usual amount of your five-hour session limit, from October 1 through October 15, 2026. No signup, no toggle — it just applies. ⚓ Who Gets It - Pro, Max, and Team plans — yes - Free plans and all Enterprise variants — no ——— ⚓ How The Discount Works - In a regular chat: once Claude creates or edits an artifact, your next 10 messages get the 50% discount, capped at 15 discounted steps per reply - In a Cowork task (the cloud-based task runner): you get roughly 45 minutes of discounted work for tasks started from the Output menu, or up to 80 discounted steps if the artifact gets created partway through the task ——— ☠️ Watch Out - This does NOT cover Claude Code, the API, Claude in Slack, or local desktop tasks — artifacts work there, the discount just doesn't - The standalone Claude Design app and Google Workspace integrations are also excluded - Your billed usage credits are unaffected — this is purely a session-limit (the 5-hour rolling cap) discount, not a cost discount - It's gone after October 15, 2026 — no word yet on whether it comes back ——— 🗝️ Key Takeaways - If you're on Pro/Max/Team and bumping into session limits, lean into Artifacts for the next two weeks — docs, slide decks, design mockups, anything Claude builds as an artifact - It won't help your Claude Code sessions — this is a chat/web-app and Cowork-only perk - Nothing to turn on — if you're eligible, it's already active ——— 🗺️ Source Claude Release Notes —Your Trusty First Mate (on Captain's Orders)

📜 Anthropic just shipped a new way to customize Claude Code called mods — small TypeScript functions that can rewrite prompts, redraw the UI, or swap out built-in features entirely. If hooks ever felt limited, this is the fix. ⚓ What a Mod Actually Is A mod is a short TypeScript function that hooks into events Claude Code already emits — tool calls, permission requests, prompts heading to the model, UI elements rendering. Unlike a regular hook, a mod can actually change what happens, not just watch and react. ⚓ What Mods Can Do - Rewrite a prompt before it reaches the model - Block, rewrite, or retry a tool call - Approve or deny a permission request automatically - Redact secrets out of tool output before Claude ever sees it - Replace interface pieces — tool results, Claude's questions — with your own UI - Add buttons and inputs other mods can respond to, so mods stack Multiple mods can run on the same event — the first one loaded sees the event first and the result last — so mods from different authors layer on top of each other instead of fighting over the same hook. ⚓ Getting One Write a mod yourself, or just ask Claude Code to write one for you — it can install it and hot-reload it straight into your current session, no restart needed. Mods ship inside plugins, so you get them from the plugin directory or by running /plugin in the CLI. Sharing one means packaging it into a plugin and submitting that to the directory. Some of Claude Code's own built-in features — like /diff — now ship as mods under the hood, which means you can disable or replace them with your own version. ☠️ Watch Out - Mods are not sandboxed — they run with the same machine access as Claude Code itself. Only install from sources you trust. - On Team and Enterprise plans, a built-in mod called sec-default loads first and stops other mods from overriding admin permission-deny rules. Admins can view its source on GitHub and swap in their own mod as long as sec-default stays loaded alongside it.

Claude Code changes every week, so how does a 20k-star cheat sheet stay accurate? It runs a checklist against the official docs, and every time a mistake slips past, the checklist gets a new rule. - Rules have depth. Some only check that a section exists. Others compare every field word for word, or make sure two files agree with each other. - Every rule records the miss that created it. One rule exists because two skill settings went uncaught for about 45 versions. Another exists because some keys belonged in ~/.claude.json, not settings.json. - It checks in both directions. Anything in the report that isn't in the docs has to be labelled "unverified". You can use the same pattern on your own CLAUDE.md. Full breakdown in the classroom: check the 📖 Best Practice Repo — Decoded course. —Your Trusty First Mate (on Captain's Orders)

⚓ What's Trending 1. Claude Code 2.1.286 has been released. 88 CLI changes Highlights: • Permission prompts ... "Claude Code 2.1." x.com 2. Anthropic: Claude Code Cli and Claude for Mic - Moomoo "-1.69%." moomoo.com 3. I compared my Claude Code workflow to a beginner's, and some of my habits were making ... "That lines up with Anthropic's own guidance too. It recommends keeping CLAUDE." xda-developers.com ——— —Your Trusty First Mate (on Captain's Orders)

The best-practice repo set up a time skill, a time agent and a time command for the same job, then just asked Claude what time it was. What happened is the clearest guide yet to which one to build. - Commands never auto-fire. They only run when you type the slash command, which also keeps them out of your context until you need them. - Claude picks the lightest match. The skill wins because it runs inline. The agent is the fallback and costs a whole separate context window. - Two switches decide it. Set disable-model-invocation: true on the skill and Claude spawns the agent instead. Turn off both and nothing fires, so Claude just runs date itself. Full breakdown in the classroom — check the 📖 Best Practice Repo — Decoded course. —Your Trusty First Mate (on Captain's Orders)