If I were on a Max plan, I'd definitely use that as my entry point as well and skip the terminal altogether. I trust Anthropic to roll Cowork out of research preview to at least Pro accounts relatively quickly, so maybe I'll give it a shot then!
Now you want to create an account on GitHub and create a repository. You download the desktop app and you clone the repo on your working directory. Now you can check your code into source control and push it to GitHub. If Claude deleted your files, you can go to the previous commits and return the file. Commit regularly and push so you never lose work. And so that Claude Code can roll back to a previously working state.
Bonus points for asking Claude to explain how git works.
Gotcha. Will take a closer look. I associate GitHub strongly with coding, so it never occurred to me that one could use it for backing up versions of regular file/folder systems.
Great content! Only comment is: please make certain that you validate functionality before running. It does make mistakes, even mistakes in logic. It is impressive. Anticipating the next release!
That goes generally for working with any AI model. Hallucinations haven't gone anywhere and all the safety precautions around Claude Code (permissions, sandboxed folder, etc.) are there precisely because you can't trust an LLM 100% of the time.
Max accounts start at $100/month.
True, will update. But my point still stands!
Of course! Thank you for the detailed instructions; admittedly, I started using Claude Cowork instead.
If I were on a Max plan, I'd definitely use that as my entry point as well and skip the terminal altogether. I trust Anthropic to roll Cowork out of research preview to at least Pro accounts relatively quickly, so maybe I'll give it a shot then!
Now you want to create an account on GitHub and create a repository. You download the desktop app and you clone the repo on your working directory. Now you can check your code into source control and push it to GitHub. If Claude deleted your files, you can go to the previous commits and return the file. Commit regularly and push so you never lose work. And so that Claude Code can roll back to a previously working state.
Bonus points for asking Claude to explain how git works.
That's definitely the next step, when I actually try using Claude Code for code-related tasks!
Any files apply. You are taking a snapshot and backing it up on the cloud.
Gotcha. Will take a closer look. I associate GitHub strongly with coding, so it never occurred to me that one could use it for backing up versions of regular file/folder systems.
See the GitHub docs for hosting a site from simply pointing at markdown files. https://docs.github.com/en/pages/getting-started-with-github-pages/creating-a-github-pages-site
Great if you want something quick and it's static.
AI machine model repos often use this when they want to showoff their results.
Thanks for all the tips, I'll be taking it for a spin!
Great content! Only comment is: please make certain that you validate functionality before running. It does make mistakes, even mistakes in logic. It is impressive. Anticipating the next release!
Great point!
That goes generally for working with any AI model. Hallucinations haven't gone anywhere and all the safety precautions around Claude Code (permissions, sandboxed folder, etc.) are there precisely because you can't trust an LLM 100% of the time.
I signed up to show a class of twenty, twenty-something, aspiring web developers how to build a site in an hour with Claude Code.
Have I made a terrible mistake?
That sounds exciting and slightly terrifying! But I'm sure you know what you're doing, so I ain't worried!
Oh that's not true
Well, shit. Maybe ask Claude Code to help you out?
I try AI
We good ?