Hello Full Stack PMs!
Welcome to Full Stack Friday, serving up the hottest AI developments fresh off the griddle. We've got 2,612 new stackers since the last send (crossed 30K!) – welcome to the stack! 🥞
Big news on my end this week – I officially opened applications for CC for PMs: Mastery. On the AI front, two trends worth paying attention to: CLIs are quietly becoming the default way agents connect to your tools, and your AI tools are starting to run without you.
Here's what we're covering:
My stuff – CC for PMs: Mastery applications are open. What it is and how to apply. And a podcast appearance.
CLIs are back – MCPs got all the hype, but CLIs are winning. Google shipped one before they even have an MCP.
Your tools went autonomous – Cursor and Claude both shipped "run while you sleep" features in the same week.
Let's do this.
📣 The Best Way to Use Claude Code
The sponsor for this newsletter is a product I actually use every single day:
Nimbalyst: The best way to use Claude Code (it’s also free!)
Nimbalyst gives Claude Code a UI, giving it some serious upgrades:
Markdown: Editing raw markdown is terrible, Nimbalyst gives you a nicely formatted, directly editable markdown files.
Mockups: Unlike Figma Make, viewing designs is not separate. You can annotate and iterate on the designs with CC directly in Nimbalyst.
Excalidraw: Make drawings with CC, edit yourself, and build working prototypes directly from the drawings.
And, again, it's completely free! My Claude Code for PMs and Claude Code for Everyone courses can be completed in about 5 hours and work perfectly in Nimbalyst.
If you already use Claude Code and want a nicer UI, or if you haven’t Claude Code because you’re intimidated by the terminal, Nimbalyst is genuinely great.
🍳 Fresh Off the Griddle
Claude Code for PMs: Mastery is here. In case you missed it, I officially opened applications for a huge upgrade to Claude Code for PMs.
6 all-new interactive modules
Advanced Claude Code techniques applied to the most common PM work like – research, building, data analysis, documents
Includes a private Slack community and weekly office hours with me.
I'm selecting 50 members for the founding class at special $200 pricing.
This is NOT first-come-first-serve – I'm reviewing applications to make sure we have a good mix, so there's still time to apply.
I'll select candidates next week
We start the week of 3/23
I'm scaling slowly from there to keep everything high quality
Advanced CC for Everyone is coming soon!
Apply here (2 mins): https://forms.gle/mG1YGYtqc6n6dgBN9
The biggest question I've been getting: should I still do the current free course? YES. Mastery is fully additive to the existing ccforpms.com course, and that course will stay 100% free forever. If you haven't done it yet, now is the perfect time so you're ready for Mastery when it starts.
CC for Everyone folks – an equally massive upgrade for you is coming soon after! Same note applies: if you haven't done the free course at ccforeveryone.com, now is great.
If you have ANY questions about any of this, just reply to this email. Happy to answer.
Cursor for Product Managers – Live session WITH Cursor: I am hosting an official live workshop with Cursor next week called “Cursor for Product Managers.” We’ll run through what a Cursor-powered PM’s day looks like from the start of a project to building a working prototype.
Will be fully hands-on and interactive!
When: Wednesday, March 11: 10-11am PST.
Sign up here (100% free): https://luma.com/vmzlmhab?tk=JIrwxu
🔧 CLIs Are Becoming the New AI Default
MCPs have been getting all the attention as the way to connect AI to your tools. But there's a quieter trend picking up: CLIs (command line interface) tools are becoming the more practical option, especially now that agents run directly on your machine.
Why they work so well for agents: CLIs are just programs installed on your computer. The agent doesn't need to hold anything in memory or maintain a connection – it calls the CLI, gets the result, moves on. LLMs are already very good at using command-line tools. And there's no MCP configuration, and much less setup friction. You install it, log in, and it works.
CLIs are ancient technology. Like, decades old. But now that agents work locally on your machine, they're making a real comeback.
The biggest signal this week: Google released an official Workspace CLI – and they don't even have an official MCP yet! And it's specifically aimed at agents.
Here’s the best video I’ve found for setting up the configuration piece. This is for the MCP but it’s the same steps for the CLI.
My take
Here's a practical tip: next time you need to do something with an external tool, ask your AI if there's a CLI for it. There often is. And it's way easier to set up than configuring an MCP.
I was kind of ahead of this trend – the Vibe Coding module in both CC for PMs and CC for Everyone uses the Vercel CLI and GitHub CLIs. That's why that module is such a delight.
Expect a lot more CLI tools from major companies to power up Claude Code and every tool like it. The simpler the interface between agents and your tools, the more useful everything becomes. I'm personally excited about this direction.
🤖 AI Is Starting to Run Without You
OpenClaw's explosion continues, reaching 250K GitHub stars this week (passing React!). One of its key features was "cron jobs" – the agent just runs tasks on a schedule without you prompting it.
Now the bigger players are building that same functionality into their products:
Cursor Automations (March 5) – Always-on agents triggered by events like GitHub pushes, Slack messages, or schedules. They spin up in a cloud sandbox and can learn from past runs. Cursor says they're running hundreds of automations per hour already.
Cursor Cloud Agents (Feb 24) – Autonomous agents on isolated VMs that build, test their own work, and open PRs. 30% of Cursor's own merged PRs are now from agents.
Claude Cowork Scheduled Tasks (rolling out Feb-March) – Write a prompt once, pick a cadence (daily, weekly, monthly), Claude runs it automatically. No code, no APIs. One limitation: desktop app needs to be open.
My take
The question to ask yourself: what would it be helpful to have an agent doing every day?
Right now it's best for monitoring, status updates, and first drafts. Imagine opening your laptop and your standup notes are already written, your metrics are already pulled, your inbox is already triaged. That's where we’re at.
I will say I honestly think this agent stuff is pretty overhyped right now – PMs, for example, don't really need mass amounts of AI slop and monitoring a few things isn't that hard. But over time, it won't just be gathering context – agents will actually take action and move things forward before you even look at them.
Expect a lot of agents content in The Full Stack PM as this evolves.
But critically: the way you set up all of this is through tools like Claude Code and Cursor. Mastering (😉) those tools is your highest leverage move right now, and will let you ride this wave instead of getting caught up in it.
😂 Meme of the Week
This is basically me:
🥓 Sizzle Reel
Other stuff worth knowing this week. This is not really a “news” newsletter – these are things you can play with right now:
GPT-5.4 – OpenAI's new flagship. 1M token context window (their largest ever), native computer use (first for a general-purpose OpenAI model), 33% fewer factual errors vs GPT-5.2.
Gemini 3.1 Flash-Lite – 2.5x faster than Gemini 2.5 Flash at $0.25 per million input tokens. Has adaptive "thinking levels" so you can dial reasoning depth per task. Cheapest frontier-adjacent model out there right now.
Claude Code Remote Control – Continue your sessions from your phone! Just run /remote-control in your Claude Code session.
🥞 The Last Pancake
If you haven't done any of my 100% free courses – what are you waiting for?! Module 1 of any of these take about an hour and will massively unlock you like it has for thousands of others already.
If you have any questions about these or CC4PMs: Mastery, feel free to reply directly to this email. I (really really try my best to) respond to every message.
Keep building,
Carl 🥞