Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

What makes Claude Code so damn good

minusx.ai

Post reply on HN
Print view View original

101–110 of 322 posts

Re: What makes Claude Code so damn good

#101

Post by CuriouslyC »

alex1138 wrote: post #5

What do people think of Google's Gemini (Pro?) compared to Claude for code? I really like a lot of what Google produces, but they can't seem to keep a product that they don't shut down and they can be pretty ham-fisted, both with corporate control (Chrome and corrupt practices) and censorship

Gemini is amazing for taking a merge file of your whole repo, dropping it in there, and chatting about stuff. The level of whole codebase understanding is unreal, and it can do some amazing architectural planning assistance. Claude is nowhere near able to do that.

My tactic is to work with Gemini to build a dense summary of the project and create a high level plan of action, then take that to gpt5 and have it try to improve the plan, and convert it to a hyper detailed workflow xml document laying out all the steps to implement the plan, which I then hand to claude.

This avoids pretty much all of Claude's unplanned bumbling.

Re: What makes Claude Code so damn good

#102

Post by dpe82 »

ComputerGuru wrote: post #99
1zael wrote: post #82

I've literally built the entire MVP of my startup on Claude Code and now have paying customers. I've got an existential worry that I'm going to have a SEV incident that will trigger a house of falling cards, but until then I'm constantly leveraging Claude for fixing security vulnerabilities, implementing test-driven-development, and planning out the software architecture in accordance with my long-term product roadma…

> but until then I'm constantly leveraging Claude for fixing security vulnerabilities That it authored in the first place?

Do you ever fix your own bugs?

Re: What makes Claude Code so damn good

#103

Post by the_mitsuhiko »

CuriouslyC wrote: post #98
the_mitsuhiko wrote: post #53

Unfortunately, Claude Code is not open source, but there are some tools to better figure out how it is working. If you are really interested in how it works, I strongly recommend looking at Claude Trace: https://github.com/badlogic/lemmy/tree/main/apps/claude-trac... It dumps out a JSON file as well as a very nicely formatted HTML file that shows you every single tool and all the prompts that were used for a session.

https://github.com/anthropics/claude-code You can see the system prompts too. It's all how the base model has been trained to break tasks into discrete steps and work through them patiently, with some robustness to failure cases.

> https://github.com/anthropics/claude-code

That repository does not contain the code. It's just used for the issue tracker and some example hooks.

Re: What makes Claude Code so damn good

#104

Post by ahmedhawas123 »

Thanks for sharing this. At a time where this is a rush towards multi-agent systems, this is helpful to see how an LLM-first organization is going after it. Lots of the design aspects here are things I experiment with day to day so it's good to see others use it as well

A few takeaways for me from this (1) Long prompts are good - and don't forget basic things like explaining in the prompt what the tool is, how to help the user, etc (2) Tool calling is basic af; you need more context (when to use, when not to use, etc) (3) Using messages as the state of the memory for the system is OK; i've thought about fancy ways (e.g., persisting dataframes, parsing variables between steps, etc, but seems like as context windows grow, messages should be ok)

Re: What makes Claude Code so damn good

#105

Post by triyambakam »

sdsd wrote: post #33

Oof, this comes at a hard moment in my Claude Code usage. I'm trying to have it help me debug some Elastic issues on Security Onion but after a few minutes it spits out a zillion lines of obfuscated JS and says: Error: kill EPERM at process.kill (node:internal/process/per_thread:226:13) at Ba2 (file:///usr/local/lib/node_modules/@anthropic-ai/claude-code/cli.js:506:19791) at file:///usr/local/lib/node_modules/@anthro…

I would try upgrading or wiping away your current install and re-installing it. There might be some cached files somewhere that are in a bad state. At least that's what fixed it for me when I recently came across something similar.

Re: What makes Claude Code so damn good

#106

Post by CuriouslyC »

the_mitsuhiko wrote: post #103
CuriouslyC wrote: post #98

Earlier quoted context omitted.

https://github.com/anthropics/claude-code You can see the system prompts too. It's all how the base model has been trained to break tasks into discrete steps and work through them patiently, with some robustness to failure cases.

> https://github.com/anthropics/claude-code That repository does not contain the code. It's just used for the issue tracker and some example hooks.

It's a javascript app that gets installed on your local system...

Re: What makes Claude Code so damn good

#107

Post by kroaton »

conception wrote: post #85

Ive seen context forge has a way to use hooks to keep CC going after context condensing. Are there any other patterns or tools people are using with CC to keep it on task, with current context until it has a validated completion of its task? I feel like we have all these tools separately but nothing brings it all together and also isn’t crazy buggy.

Load up the context with your information + task list (broken down into phases). Have Sonnet implement phase one tasks and mark phase 1 as done. Go into planning mode, have Opus review the work (you should ideally also review it at this point). Double press escape and go back to the point in the conversation where you loaded up the context with your information + task list. Tell it to do phase 2. Repeat until you run out of usage.

Re: What makes Claude Code so damn good

#108

Post by kroaton »

kroaton wrote: post #107
conception wrote: post #85

Ive seen context forge has a way to use hooks to keep CC going after context condensing. Are there any other patterns or tools people are using with CC to keep it on task, with current context until it has a validated completion of its task? I feel like we have all these tools separately but nothing brings it all together and also isn’t crazy buggy.

Load up the context with your information + task list (broken down into phases). Have Sonnet implement phase one tasks and mark phase 1 as done. Go into planning mode, have Opus review the work (you should ideally also review it at this point). Double press escape and go back to the point in the conversation where you loaded up the context with your information + task list. Tell it to do phase 2. Repeat until you run…

From time to time, go into Opus planning mode, have it review your entire codebase and tell it to go file by file and look for bugs, security issues, logical problems, etc. Have it make a list. Then load up the context + task list...

Re: What makes Claude Code so damn good

#110

Post by koakuma-chan »

the_mitsuhiko wrote: post #103
CuriouslyC wrote: post #98

Earlier quoted context omitted.

https://github.com/anthropics/claude-code You can see the system prompts too. It's all how the base model has been trained to break tasks into discrete steps and work through them patiently, with some robustness to failure cases.

> https://github.com/anthropics/claude-code That repository does not contain the code. It's just used for the issue tracker and some example hooks.

https://github.com/dnakov/claude-code :trollface:
Post reply on HN