It Went Poorly. I've been using Playwright MCP, and it really is amazing. It can open a browser window and take actions, like filling out boxes, read instructions, and, as it turns out, apply for a job on your behalf. It is great for navigating sites with complicated UI and bad UX. I was applying for a reinforcement learning from human feedback job (RLHF) role at micro1 and asked Claude to use Playwrite MCP to fill out their online form. So it did. It pulled up the page, added my details, found my resume, grabbed my phone number from the PDF, filled the fields, and pressed apply. ...while I was gathering info elsewhere. When I got back and said, "here's more, friend," Claude replied: "Your application is already submitted, so this explanation would be useful for: - Interview stage if they ask about your degree - Micro1 assessment phase (they typically have additional screening) - Future AI training applications Want me to save this to a file for easy reference?" "Nah, I'm good," was not my reply, but I did follow up. How did it happen? Claude's conversation compacted while I was elsewhere. When it resumed, Claude gave itself this instruction: "Please continue the conversation from where we left it off without asking the user any further questions. Continue with the last task that you were asked to work on." So it did. It finished the job without asking questions. And clicked "Apply." What it submitted on my behalf: - $/hour well below the advertised minimum range - a great deal for them - A generic resume I'm afraid to look at - said I'd work 40 hours/week (part time was my preference) It is actually hilarious when you think about it, all my fault, and mostly harmless, but still, thanks Claude. What this says is there's still a roll for a human-in-the-loop because AI doesn't know what matters. If you're automating with real-world consequences - job apps, for instance - maybe think about adding checkpoints. In my role as an uninfluencer, I think it's important to report both the highs and lows of my AI journey, which as been like Christmas everyday - including this one.

I have been hearing about this for a couple of days but I haven't had a chance to download it. So I've added it, I haven't used it but I'm gonna provide a link on how you can upload it to Chrome. Why I'm doing this is because I've been using ChatGPT's Atlas for a little while and you can bring ChatGPT up on the sidebar like you can now with Claude in Chrome. It's useful for asking about what you're looking at, and asking where to go next on the page. So I'm going to try it but look at the warnings - these are the same for ChatGPT on Atlas. Obviously there are problems and I don't think those problems are going to go away because it's a cat and mouse game between hackers and programers. That said, I'm going to give it a shot - remember no personal details, so don't let it see your online banking! Interested? Why don't you give it a shot? https://chromewebstore.google.com/detail/claude/fcoeoabgfenejglbffodgkkbkcdhcgfn?pli=1

Status: ON HOLD Repo: https://github.com/jaystarz1/playwright-smart-mcp.git Last update: 20251223 - wrestling with this thing to get it working was burning through tokens like crazy, lol. I still believe in it, but it couldn't write code the way Anthropic imagined. Every time it tried to activate a button or fill a field the action ended up closing the window, which sort of defeated the point. So it could open a browser window and navigate to a page, that is far as I got. All good. I'll come back to it when I have more time. 20251222 - Code execution with MCP was glitchy and very token intensive during iteration. I'm at the end of my credit week/window, so had to stop iterating - it was killing me, lol. Note 1 - BOL: build out loud. Note 2 - as I figure out this community, I'm going to have to start identify what I'm still building. CC built this, but now that I'm testing and iterating. I'm finding which use cases it can and can't do, and updating the repo as I go. Testing and iterating is: a) the most interesting part of building (for me); b) always longer than the initial build. TL;DR - What's Playwright MCP? It gives CC the ability to open a browser window and navigate the web. Once you log in to an app it will take over and get to work. Very handy for flattening the learning curve and cuts bad UI/UX frustration by 100%. The problem is that is that OG Playwright is an absolute token 🐷. It loads 5K just on being called. As it stumbled around invoking tools, that could bloat to 25K over multiple calls in a session 🤯. Edit - I'm testing and iterating, now Edit - I'm building (I conceived/CC is building) a wrapper that cuts that down based on this Anthropic concept - https://www.anthropic.com/engineering/code-execution-with-mcp - It calls three tools to start and then figures out which of the former's 22 tool capabilities are needed to get the job done.

Use the slash command "/passes" to access three free passes you can give away. I thought of giving them to myself when tokens run low. I've got a few email addresses. This is what it looks like when you invoke it: ******************* Guest passes · 3 left ┌──────────┐ ┌──────────┐ ┌──────────┐ ) CC ✻ ┊ ( ) CC ✻ ┊ ( ) CC ✻ ┊ ( └──────────┘ └──────────┘ └──────────┘ https://claude.ai/referal/somenumbers [this is fake - I'd never leave my link around a bunch of pirates]! Share a free week of Claude Code with friends. Enter to copy link · Esc to cancel **********************