Most active commenters
  • CuriouslyC(3)

←back to thread

How to use Claude Code subagents to parallelize development

(zachwills.net)
280 points zachwills | 11 comments | 09 Sep 25 13:21 UTC | HN request time: 0.945s | source | bottom
Show context
CuriouslyC ◴[13 Sep 25 04:52 UTC] No.45229400[source]▶
As someone who's built a project in this space, this is incredibly unreliable. Subagents don't get a full system prompt (including stuff like CLAUDE.md directions) so they are flying very blind in your projects, and as such will tend to get derailed by their lack of knowledge of a project and veer into mock solutions and "let me just make a simpler solution that demonstrates X."

I advise people to only use subagents for stuff that is very compartmentalized because they're hard to monitor and prone to failure with complex codebases where agents live and die by project knowledge curated in files like CLAUDE.md. If your main Claude instance doesn't give a good handoff to a subagent, or a subagent doesn't give a good handback to the main Claude, shit will go sideways fast.

Also, don't lean on agents for refactoring. Their ability to refactor a codebase goes in the toilet pretty quickly.

replies(5): >>45229506 #>>45229671 #>>45230608 #>>45230768 #>>45230775 #
1. zarzavat ◴[13 Sep 25 05:52 UTC] No.45229671[source]▶
> Their ability to refactor a codebase goes in the toilet pretty quickly.

Very much this. I tried to get Claude to move some code from one file to another. Some of the code went missing. Some of it was modified along the way.

Humans have strategies for refactoring, e.g. "I'm going to start from the top of the file and Cut code that needs to be moved and Paste it in the new location". LLM don't have a clipboard (yet!) so they can't do this.

Claude can only reliably do this refactoring if it can keep the start and end files in context. This was a large file, so it got lost. Even then it needs direct supervision.

replies(4): >>45230552 #>>45231336 #>>45231629 #>>45232104 #
2. diggan ◴[13 Sep 25 09:02 UTC] No.45230552[source]▶
> Humans have strategies for refactoring, e.g. "I'm going to start from the top of the file and Cut code that needs to be moved and Paste it in the new location". LLM don't have a clipboard (yet!) so they can't do this.

For my own agent I have a `move_file` and `copy_file` tool with two args each, that at least GPT-OSS seems to be able to use whenever it suits, like for moving stuff around. I've seen it use it as part of refactoring as well, moving a file to one location, copying that to another, the trim both of them but different trims, seems to have worked OK.

If the agent has access to `exec_shell` or similar, I'm sure you could add `Use mv and cp if you need to move or copy files` to the system prompt to get it to use that instead, probably would work in Claude Code as well.

3. lupire ◴[13 Sep 25 11:51 UTC] No.45231336[source]▶
Remember 20 years ago when Eclipse could move a function by manipulating the AST and following references to adjust imports and callers, and it it didn't lose any code?
replies(3): >>45232292 #>>45233943 #>>45234165 #
4. wahnfrieden ◴[13 Sep 25 12:45 UTC] No.45231629[source]▶
Codex’s model is much better at actually reading large volumes of code which improves its results compared with CC
5. brookst ◴[13 Sep 25 13:56 UTC] No.45232104[source]▶
Claude’s utility really drops when any task requires a working set larger than the context window.

On the one hand, it’s kind or irritating when it goes great-great-great-fail.

On the other hand, it really enforces the best practices of small classes, small files, separation of concerns. If each unit is small enough it does great.

Unfortunately, it’s also fairly verbose and not great at recognizing that it is writing the same code over and over again, so I often find some basic file has exploded to 3000 lines, and a simple “identity repeated logic and move to functions” prompt shrinks it to 500 lines.

6. Yeroc ◴[13 Sep 25 14:21 UTC] No.45232292[source]▶
I think it's likely that these agent-based development will inevitably add more imperative tools to their arsenal to lower cost, improve speed and accuracy.
7. mleo ◴[13 Sep 25 17:50 UTC] No.45233943[source]▶
It’s still early days for these agents. There isn’t any reason the agents won’t build or understand AST in the future to more quickly refactor.
replies(1): >>45234170 #
8. CuriouslyC ◴[13 Sep 25 18:17 UTC] No.45234165[source]▶
I have a suite of agent tools that is just waiting on my search service for a release, it includes `srefactor` and `spatch` commands that have fuzzy semantic alignment with strong error guards, they use LSP and tree sitter to enable refactoring/patching without line numbers or anything and ensure the patch is correct.
replies(1): >>45234543 #
9. CuriouslyC ◴[13 Sep 25 18:18 UTC] No.45234170{3}[source]▶
Why do the agents need to build or understand it? Just give them tools to work with it like we would.
replies(1): >>45234288 #
10. LtdJorge ◴[13 Sep 25 18:34 UTC] No.45234288{4}[source]▶
Everyone talking about MCP and they haven’t figured this out. Actually, JetBrains has an IDE MCP server plugin, although I haven’t tried it.
11. catlifeonmars ◴[13 Sep 25 19:13 UTC] No.45234543{3}[source]▶
Nice. This sounds like the right approach. As an aside, it’s crazy that a mature LSP server is not a first class requirement for language choice in 2025. I used to write mini LSP servers before working on a project starting when LSP came out a few years ago. Now that there is wider adoption, I don’t find myself reaching for this quite as often, but it’s still a really nice way to ease development on mature codebases that have grown their own design patterns.