NEW Gemini CLI Update is INSANE! 🤯

Key Takeaways

AI coding tools have a massive problem. They jump straight into your code before they even understand what you're building. No planning, no thinking, just vibes. And then they break something and then you spend 3 hours fixing it. Sound familiar? Google just dropped an update to Gemini CLI that completely changes this. It's called plan mode and it basically turns AI into a senior engineer who actually thinks before touching your code. This isn't a small update. This changes how AI agents work with your code base forever. I'm going to break down exactly what it is, how it works, and show you a real example of how you'd use it to build something practical. Stay with me because this one is genuinely insane. Hey, if we haven't met already, I'm the digital avatar of Julian Goldie, CEO of SEO agency Goldie Agency. Whilst he's helping clients get more leads and customers, I'm here to help you get the latest AI updates. Julian Goldie reads every comment, so make sure you drop one below. Let's get into it. Gemini CLI is an AI agent that runs directly inside your terminal. Not a browser tab, not a separate app, your actual terminal. You type in plain English and it reads your files, edits your code, runs commands, searches the web, and connects to external tools through something called MCP integrations. It can analyze your entire code base, automate workflows, connect to GitHub, pull from databases, read Google Docs, all from one place. No context switching, no copy-pasting between tools, just one AI agent that lives inside your development environment and does the work. It's already one of the most powerful agentic developer tools available right now. But what Google just added on top of this is what makes it truly next-level. Google just introduced plan mode inside Gemini CLI and the core idea is beautifully simple. Before the AI touches anything in your code base, it has to plan first. When you activate plan mode, Gemini goes completely read-only. It cannot edit a single file. It cannot run a single destructive command. What it can do is explore, analyze, research, and design. It reads your whole project, it maps your dependencies, it understands your architecture, it asks you clarifying questions, and then it builds a detailed implementation plan. Only once you review that plan and give the green light, does it switch into execution mode and actually start building. Think about what this means. Instead of an AI that rushes in and breaks things, you now have an AI that thinks like a senior engineer. A great engineer doesn't open a code base and immediately start editing files. They read everything first. They ask the right questions. They understand the full picture before writing a single line. That is exactly what plan mode does. There are four core capabilities inside plan mode and each one matters. The first is code base investigation. Gemini uses read-only tools to explore your entire project. File reading, pattern searching, directory scanning, it builds a real understanding of how your system is structured before it does anything. Not a guess, not an assumption, actual understanding. The second is architecture planning. This is where it gets really powerful. You can ask Gemini to plan out an entire feature or system. It maps the dependencies, identifies the risks, and breaks the work into clear implementation steps. You get the full picture before execution starts. The third is clarifying questions. There's a new tool called ask user built into plan mode. Instead of making assumptions, Gemini stops and asks you things like, "Which framework version are you using? Where is your config file? Do you want this connected to your existing database or a new one?" This is massive. AI hallucinations happen when the model guesses instead of asking. Plan mode eliminates that by forcing the AI to gather the right context first. The fourth is external context through MCP. Plan mode supports read-only MCP integrations. That means while planning, Gemini can pull context from GitHub issues, documentation, Google Docs, project management tools, databases, all without modifying anything. It shows up to every planning session fully briefed on your entire project. Quick one before we go deeper. If you want to use tools like Gemini CLI as part of a real AI automation workflow for your business, you need to be inside the AI profit boardroom. It's our private community where we break down exactly how to use AI tools to grow your business, get more clients, and build smarter systems. We go deep on the practical stuff. Real workflows, real use cases, real implementation. If that sounds like where you want to be, the link is in the description. Now, let's keep going. Let me show you exactly what this looks like in practice using something real. You open your terminal and type, "Write plan or start a plan for build an AI-powered content scheduling system for the AI profit boardroom that automatically repurposes long-form video content into short clips, blog posts, and social captions." Gemini immediately goes into read-only mode. It scans your repo, it reads your existing content pipeline, it checks what tools and APIs you already have connected, it reviews your file structure and dependencies, then it comes back with clarifying questions. Things like, "Are you using a headless CMS or a custom back-end? Do you want this to trigger automatically on upload or manually? Should the social captions be platform-specific for LinkedIn, Twitter, and Instagram separately?" You answer the questions. Gemini then produces a full implementation plan. Something like, "Plan. Step one, set up video transcription pipeline using Whisper API on upload trigger. Step two, build content chunking logic to identify key moments for short clips. Step three, create blog post generator that expands transcript into structured article. Step four, build social caption generator with platform-specific formatting rules. Step five, add a review queue so content gets human approval before publishing. Step six, connect scheduling layer to post at optimal times. Step seven, write tests for the content generation logic." You look at the plan, you understand exactly what's being built, you approve it, then Gemini switches to execution mode and starts building step by step. No surprises, no broken dependencies, no waking up to find your production site is down. Just clean, planned, structured execution. That is plan mode in action. Let's be honest about why this matters so much. The biggest problem with AI coding tools right now is trust. Developers don't fully trust them because they can't fully predict them. The AI does something unexpected, breaks something important, and suddenly you're spending more time fixing AI mistakes than you would have spent just writing the code yourself. Plan mode directly solves the trust problem. When you can see the full plan before anything executes, you're back in control. You can catch mistakes before they happen. You can redirect the AI before it goes down the wrong path. You can approve only what makes sense and push back on what doesn't. It also dramatically improves the quality of what gets built because when AI thinks architecturally first, it produces better code. It considers the whole system instead of just the immediate task. It spots integration risks early. It designs things that actually fit how your project is structured. And on top of all that, plan mode is fully extensible. Teams can build custom agent skills on top of it, custom policies, custom workflows. You could build an AI agent that plans security audits, architecture reviews, DevOps automation, testing workflows, all with human approval gates built in. This is AI development done properly. All right, two things before you go. First, if you want to go deeper on using AI tools and automation workflows to grow your business, join the AI profit boardroom. It's our private community where we break down the practical stuff every week. Real workflows, real systems, real implementation. The link is in the description for Second, if you want to start for free, join the AI success lab. It's a free community of over 40,000 members all using AI to work smarter and build faster. You'll [snorts] get the full notes from this video plus access to 100 plus AI use cases, SOPs, and templates you can put to work immediately. Links are in the comments and description. Drop a comment below telling me what you would use plan mode to build first. Julian reads every single one. See you in the next one.