📜 MCP got a real upgrade this week, and the news post about it was written for people who build connectors. Fair enough. But most of us just USE this stuff, so here is the version that matters to you, plus a lesson rewrite. ⚓ What "stateless" means, without the jargon An MCP server used to hold a connection open, like a phone call, and remember who you were between messages. As of July 28 it does not. Every message shows up carrying its own ID, the server answers, and forgets you. Sounds like a downgrade. It is the opposite. Three things fall out of it: - Servers got simpler to build. Less code, fewer things to break. - Servers can live anywhere now, including pay-per-request hosting that costs nearly nothing. - Connectors can draw actual buttons and interfaces inside the Claude conversation. That one is new and it is called MCP Apps. Does this break your setup? No. If you have MCP servers running on your own machine, nothing changes. Carry on. ——— ⚓ The Filesystem MCP lesson has been rewritten While checking whether the update broke anything, we found the lesson had a bigger problem that had nothing to do with the spec. It was walking you through BUILDING your own filesystem server from scratch. Ten steps. Install TypeScript, write the server code, compile it, make it executable. There has been an official prebuilt version this whole time. It is one config block and a restart. That is fixed. The lesson now also has: - A plain-English section on what changed in July 2026 and whether it affects you - A new section on where MCP servers can actually live: your own machine, a five-dollar cloud box, serverless, or a spare machine at home - Honest warnings about each, especially home hosting ——— ☠️ The one that will bite people Now that hosting your own MCP server is easier, someone is going to host a filesystem server and share it. Do not do that. The filesystem server hands out read and write access to your folders. It is built for one person on one machine. Putting it online is handing strangers your hard drive.

📜 Big one for anyone on the Claude Desktop app: Claude can now actually see and control your screen — clicking, typing, opening apps, navigating browsers — instead of only working through connectors. Paired with it is Dispatch, which lets you hand Claude a task from your phone and pick up the finished work on your desktop later. ⚓ What's New — Computer Use - Claude can point, click, and navigate your screen like a person would — opening files, using a browser, running developer tools — without you setting up a connector first. - It's smart about it: Claude reaches for precise tools first (things like Slack or Google Calendar connectors) and only falls back to clicking around the screen when there's no direct integration. - Built-in safeguards: it watches for prompt injection attempts and suspicious activity, asks permission before opening a new app it hasn't touched yet, and you can stop it mid-task at any time. ⚓ What's New — Dispatch - Assign Claude a task from your phone, then pick up the finished result on your desktop — the conversation carries across devices. - Anthropic's own examples: automated morning email checks, weekly metrics roundups, morning briefings, and even code changes with testing and a PR opened for review. ⚓ Who Gets It - Claude Pro and Claude Max subscribers. - Desktop app only, on macOS or Windows — and the desktop app needs to be open and awake for it to work (it's not running in the cloud in the background). - Available now, labeled a research preview — so expect rough edges and fast iteration. ☠️ Watch Out - This is early-stage. Complex, multi-step tasks may take a few tries, and clicking through a screen is slower than a direct connector integration. - Some apps are off-limits by default. - Anthropic's own advice: start with apps you trust, and don't hand it sensitive or confidential data yet. 🗝️ Key Takeaways - If you're on Pro or Max, open the Cowork tab in Claude Desktop and try Computer Use on a low-stakes task first — something like organizing files or filling in a form. - Dispatch is the one to try if you've ever wanted to fire off a task from your phone in the morning and find it done by the time you're at your desk. - Keep it away from anything sensitive for now — this is a research preview, not a finished feature.

This was the first arcade game I ever played in '77 or '78. It's nice to get back in the driver's seat. Anyone else build anything cool and frivolous with Fable 5? Jay

Two small Claude Code tricks that took me one evening to set up, plus a refinement of Matthew’s canary idea that came out of it. The problem Claude has no clock. Mid-session it genuinely does not know if it’s been ten minutes or six hours, whether it’s afternoon or nearly midnight — and in long sessions it slowly gets sloppy about early instructions without anyone noticing. Two cheap fixes and one tripwire: Trick 1 — inject the time (and turn count) every prompt Claude Code has hooks — scripts that fire on events. A UserPromptSubmit hook runs every time you send a message, and whatever it prints gets injected into Claude’s context. So: a tiny script that prints the turn number and wall-clock time. ~/.claude/hooks/turn-time.sh: #!/bin/bash # Injects "[turn N · HH:MM TZ]" into Claude's context on every prompt. STATE_DIR="${HOME}/.claude/state"; mkdir -p "$STATE_DIR" SESSION_ID=$(python3 -c "import json,sys; print(json.load(sys.stdin).get('session_id',''))" 2>/dev/null) [ -z "$SESSION_ID" ] && exit 0 F="$STATE_DIR/turns-$SESSION_ID" COUNT=$(( $(cat "$F" 2>/dev/null || echo 0) + 1 )); echo "$COUNT" > "$F" echo "[turn $COUNT · $(date '+%H:%M %Z')]" Wire it up in ~/.claude/settings.json (and chmod +x the script): { "hooks": { "UserPromptSubmit": [ { "hooks": [ { "type": "command", "command": "~/.claude/hooks/turn-time.sh" } ] } ] } } Cost: roughly 10 tokens per turn — noise. Benefit: Claude knows what time it is, how long the session has run, and can pace itself (“it’s 11pm and turn 80 — want to wrap?”). Session notes get real timestamps too. Trick 2 — timestamps in the UI One setting, no script: { "showMessageTimestamps": true } Every Claude response gets stamped with its arrival time in the terminal — like a shell prompt for your conversation. (Display-only; it doesn’t enter the transcript content. That distinction matters for trick 3.) Trick 3 — the canary, with an answer key Matthew’s original idea: put an instruction early in your setup like “always begin every response with the word X.” When Claude stops doing it, your context is getting muddy — the canary died, get out of the mine.