How I use Claude Cowork (Senior Software Engineer Tips)

This video is sponsored by HubSpot. Claude Co-work has been quietly handling some of the most tedious parts of engineering work, and most engineers still haven't touched it. Not marketers, not business owners, engineers. People who write code, ship features, and sit through design reviews. And it's more useful for that kind of work than most people realize, but probably not for the reasons you expect. Let's get into it. Hi friends, I'm Maddie. I'm a senior software engineer who previously worked at Google and interned at other big tech companies like Amazon, IBM, and Microsoft. Here's the thing. A lot of co-work demos focus on generic productivity. Summarize this, organize that, draft a doc. This can be useful, but it's not the reason I think engineers should care. The real usefulness of co-work comes when the task is recurring, pulls from multiple systems, and doesn't require you to sit there while it runs. Co-work isn't just another chat interface. You can schedule recurring tasks, connect it to your tools, and let it work asynchronously in the background while you move on to something else. Those three things together unlock a very specific category of engineering workflows. The recurring multi-source nobody wants to do a tasks that eat your time every single week. So that's what we'll dive into for this video. Four use cases built specifically around what co-work can unlock for you. Let's get into it. We'll start with some quick context, especially if you heard of Claude Code and are wondering what the difference between Claude Code and Claude Co-work is. Claude Code runs in your terminal. It's built for developers who want to write and execute code, manage builds, run tests, and iterate on a codebase with a tight feedback loop. It's powerful, but it assumes you're comfortable in a CLI, and that you're actively in the loop the whole time. Co-work is different in three specific ways. First, it has a visual desktop interface. No terminal, no slash commands, just a folder you point it at and a task you describe. Second, it connects to external tools natively. Jira, GitHub, Google Drive, Gmail, Calendar, and more through what it calls connectors. Third, it can run tasks on a schedule and in the background, meaning that you can kick something off and come back to the finished output. So, the mental model is this. Cloud code is for active development work where you need a real-time feedback loop. Codework is for everything else. file processing, document generation, connected tool workflows, and recurring tasks that need to run whether you're at your desk or not. So, here's how you start treating it less like a chat window and more like a task handoff. Define the deliverable, give it the context, and let it work. If you're just getting started with co-work and staring at a blank task field wondering how to actually use it, HubSpot put together something that helps with that. It's called the Claude Co-work Stack and it's a free guide with 12 fully built prompts that show you exactly how to get co-working for you. Each prompt comes with a use case, the inputs you need to provide, and the expected output. So, you're not just getting a prompt, you're getting a full workflow template. things like a weekly operations brief that pulls from your calendar and project folders, a meeting prep generator that researches everyone and your schedule, a research synthesis report that turns a folder of articles and notes into a coherent document, and a batch document generator for when you need to create a lot of similar things at once. The prompts lean toward business and ops use cases, so if you're engineer, you'll probably adapt them rather than use them verbatim. But seeing 12 concrete examples of how to structure co-work tasks, what context to give it, and what a good output looks like is a solid way to build intuition fast, especially when you're starting out. My favorite is the research synthesis prompt. You drop in a folder of articles or notes and get back one clean document. I suggest you take a look at the Claude Co stack. It's completely free and you can check it out in the link in the description. Thanks to HubSpot for sponsoring this video. And before you go set up co-work, there's a few items worth knowing. Workspace is a local folder you grant co-work access to. Everything it reads from and writes to lives here. There's no cloud sync or external storage. It's just your machine. Connectors are the integrations. So, for example, Jura, Google Drive, Calendar, and more. They're what lets co-work pull live data from your actual tools rather than just working with files that you manually dropped into a folder. You authorize them once in the desktop app, and they're available across all your tasks. Skills are reusable task templates you build and save. Instead of writing out the same long prompt every time, you define the task once, what inputs it needs, what it should do, what the output should look like, and it becomes a one-click workflow. Schedule tasks are skills configured to run automatically at a set time or interval. You set the schedule once and it fires on its own, daily, weekly, whatever you configure it to be. And finally, artifacts are the output files co-work creates and saves to your workspace after a task runs. for example, markdown summary, an HTML file, a reorganized folder. They persist in your workspace, so you can reference, share, or iterate on them. So, now let's talk about the use cases. If you're in tech, you probably subscribe to more newsletters than you actually read. I have a newsletter's label in Gmail that gets about 20 to 30 emails a week. AI updates, engineering blogs, job market stuff, industry news. Useful content for sure, but I never have time to sit down and read through all of it in one go. The workflow use is this. Every Sunday evening, a scheduled task connects to Gmail through the connectors feature, pulls everything from the newsletter label that came in that week, and generates a single digest file. Key takeaways from each one organized by topic. Monday morning, I've got only one document instead of 20 unread emails. What makes this better than just reading the newsletter is that co-work pulls them all at once and synthesizes across them. So instead of reading a separate email about AI model releases, another one about hiring trends, and another about new dev tools, I get one document that surfaces the through line, what actually moved this week, and what is worth paying attention to. It's a different reading experience than going one by one. You can also customize the output. Ask it to flag anything relevant your job search or anything about a specific technology you're learning. It applies that filter every week without you having to respppecify it. The setup takes about 10 minutes, connecting Gmail, pointing it at the right label, writing the output format you want, then it runs on its own. The small caveat is that this is desktop automation, so it only fires when the app is open. If your laptop's closed when it's scheduled, it actually only runs when you open it. Now, let's talk about use case two, project directory cleanup before a handoff. Every engineer has a project directory that's become a mess over time. Stale branches checked out locally, duplicate config files, outdated reads that describe a setup nobody uses anymore, built artifacts that never got cleaned up. You know it needs fixing, but you never prioritize it because it's not urgent. This is a good co-work task because it's tedious, it's well defined, and it benefits from background processing. Point co-work at the directory. Describe the structure you want, how files should be organized, what naming conventions to follow, what can be flagged for deletion versus what should be kept, and let it run while you're in a meeting. So, you can say something like, "Audit this project directory, organize files into logical subfolders, flag any files that appear to be duplicates or outdated artifacts, and generate a summary of what was moved and what needs manual review." It won't make judgment calls that require your context. It'll flag those for you rather than touch them. But it handles the mechanical layer, the renaming, the reorganizing, the folder structure. What would take you an hour of tedious clicking takes co-work a few minutes in the background. This is especially useful before a handoff. A clean, wellorganized directory is something that the next engineer will notice immediately, and it's the kind of thing that's easy to depprioritize until the last minute. The third use case is to write a post-mortem first draft from incident data. When an incident happens, the postmortem is one of those things everyone knows is valuable and no one wants to write while they're still recovering from the incident itself. Here's the workflow. After things are resolved, you can drop your incident data into a co-work folder. timeline notes, log excerpts, a summary of what happened. Connect to your instant tracking tool if you have one, and ask co-work to populate a postmortem template. So, for example, you can say something like, "Given this incident data, draft a postmortem covering the timeline, root cause, contributing factors, impact, and action items. What you get back is a structured first draft. The timeline is filled in. The obvious contributing factors are surfaced, and the action flow section has starting points. You're editing and adding judgment rather than building from scratch at the end of an already long day. The connector piece really matters here. If your incident tracking lives in a tool co-work can reach, it can pull the data directly rather than by you having to dump it manually. That's the difference between a 5-minute setup and a 20-minute one. And finally, use case 4 is turning notes and papers into interactive visual explainers. This one is less about workflows and more about learning, which for engineers is basically a permanent state. Whether you're trying to actually understand something technical, not just read about it, but internalize it, static notes only go so far. What actually happens is seeing it move. A diagram that steps through a process, a visualization where you can poke at the inputs and watch the outputs change. Something interactive that you can come back to. So, here's the workflow. Drop your notes, a paper you've been reading, or even a set of URLs into a co-work folder and ask it to generate an interactive HTML explainer. It produces a self-contained HTML file you can open in any browser with no dependencies and no setup. This is the kind of thing that would have taken a few hours to build yourself and instead is done in the background while you're doing something else. So, for example, if you're studying system design for interviews, you could ask it to generate a visual walkthrough of how a rate limiter works. Inputs, token bucket logic, what happens under load. If you're learning about how LLMs work under the hood, ask for an interactive breakdown of the attention mechanism. If you're trying to understand a codebase's request lifestyle, give it the relevant files and ask for an animated flow diagram. These aren't just pretty pictures, but rather they become personal study aids you can actually reuse. The output list in your folder, you can share it and you can iterate on it. You can ask co-work to add a section, change the framing, make it simpler, and it will update the file. For anyone actively studying for interviews or picking up a new technical area, this is one of the most underrated things you can build with co-work. And finally, there's one thing worth mentioning. Co-work also has a browser extension that lets you interact with web pages directly. So for example, filling forms, navigating a URL, or scraping pages without an API. The capability is real and quite interesting, but I treat the browser extension as experimental rather than core workflow infrastructure. It can navigate pages, but the control loop is still relatively slow and fragile compared with other native connectors. For anything important, use a connector or API based workflow first and only fall back to browser control when there is no cleaner integration. However, this is definitely worth knowing it exists and worth checking back on as it improves. And that's all I have for you in this video. Four workflows all built around what co-work really does well. Scheduling, live connectors, and background processing. If you found this helpful, hit that like button, hype the video, and subscribe. I post weekly videos on software engineering, AI tools, and career stuff. Thanks for watching and I'll see you next one.