Claude Code Checkpoints

Post by stavros » Thu, Aug 28, 2025, 9:07 PM UTC

fprotthetarball wrote: post #25

mhast wrote: post #8

Earlier quoted context omitted.

Git won't catch new files the agent is adding. To get around that you can of course always add all new files, but then you'll potentially have your repo polluted with a bunch of temporary scratch files instead. You can typically go back and edit git history. But it will require force push and breaking changes. And a few sacrifices to ensure that it doesn't make a mistake because then your repo is potentially broken.…

> Git won't catch new files the agent is adding. Another problem I inadvertently dodged by using Jujutsu with Claude Code :) I tend to send a lone "commit" message to Claude when I think I'm in a spot I may want to return to in the future, in case the current path doesn't work out. Then Claude commits it with a decent message. It knows how to use jj well enough for most things. Then it's really easy to jj new back to…

Post by bdangubic » Thu, Aug 28, 2025, 9:58 PM UTC

slavakurilyak wrote: post #94

This looks well-built, but have you considered using Jujutsu (jj) instead? It elegantly solves the snapshot/checkpoint problem without needing MCP servers or a separate app. With jj, every file change is automatically captured (no manual commits needed), and you can create lightweight "sandbox" revisions for each Claude Code task. When things go wrong, `jj undo` instantly reverts to any previous state. The operation…

Post by rubslopes » Thu, Aug 28, 2025, 11:10 PM UTC

earleybird wrote: post #104

bdangubic wrote: post #103

Earlier quoted context omitted.

I dont understand the need for this at all. I use the git. Inside one CC session on average there will be 80-160 commits. I can revert back whatever I want whenever I want. Whenever context gets hairy I clear and ask CC to review specific commits to continue or whatever. I must be missing something here completely what is the need for jj at all…

Pretty much bang on. I do have to remind myself that discipline in commit messages helps future me.

Post by CuriouslyC » Fri, Aug 29, 2025, 12:03 AM UTC

anuramat wrote: post #68

CuriouslyC wrote: post #26

Earlier quoted context omitted.

While disaster is strong language, Claude Code isn't really a well engineered product, they're just kinda trying shit, they don't have a clear long term vision. The core prompts and agent loop are good though, it's too bad it's not open source so someone could implement them in a client with good UX/engineering (at least without disassembling claude code and being legally questionable).

> not opensource You can extract prompts with mitmproxy/netcat, and AFAIK there isn't much more to it (bash and todo list are all you need in terms of tools), there's already a lot of simpler tools with better ux: - sst/opencode and charmbracelet/crush -- related "cc clones" with top tier UX; opencode has near feature parity with cc, crush is more barebones - block/goose -- a lot of multi-model features and extension…

Post by CuriouslyC » Fri, Aug 29, 2025, 1:39 AM UTC

anuramat wrote: post #108

CuriouslyC wrote: post #107

Earlier quoted context omitted.

Most of those features aren't executed well though. Hooks aren't ubiquitous and have very little transformational ability, subagents are routed poorly and lack intelligent context inheritance, etc. It is a good playground to get work done for a great price while building intuition though.

What exactly are your gripes with hooks, wdym by "transformational"? Regarding ubiquity, there's already quite a few repos with collections of hooks; I've only tried notifications and format-on-write so far, the only problem I noticed is that the provided formatting hook is confusing for the model (maybe cc could backtrack and prefill the tool call with the formatter output?) Subagents are indeed kinda useless, but i…