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

A staff engineer's journey with Claude Code

sanity.io

Post reply on HN
Print view View original

21–30 of 410 posts

Re: A staff engineer's journey with Claude Code

#21

Post by adastra22 »

swframe2 wrote: post #3

Preventing garbage just requires that you take into account the cognitive limits of the agent. For example ... 1) Don't ask for large / complex change. Ask for a plan but ask it to implement the plan in small steps and ask the model to test each step before starting the next. 2) For really complex steps, ask the model to write code to visualize the problem and solution. 3) If the model fails on a given step, ask it t…

This is why the jobs market for new grads and early career folks has dried up. A seasoned developer knows that this is how you manage work in general, and just treats the AI like they would a junior developer—and gets good results.

Re: A staff engineer's journey with Claude Code

#22

Post by NitpickLawyer »

paulcole wrote: post #10
swframe2 wrote: post #3

Preventing garbage just requires that you take into account the cognitive limits of the agent. For example ... 1) Don't ask for large / complex change. Ask for a plan but ask it to implement the plan in small steps and ask the model to test each step before starting the next. 2) For really complex steps, ask the model to write code to visualize the problem and solution. 3) If the model fails on a given step, ask it t…

> Ask for a plan but ask it to implement the plan in small steps and ask the model to test each step before starting the next. Tried this on a developer I worked with once and he just scoffed at me and pushed to prod on a Friday.

> scoffed at me and pushed to prod on a Friday.

that's the --yolo flag in cc :D

Re: A staff engineer's journey with Claude Code

#23

Post by lordnacho »

I'm using Claude all the time now. It works, and I'm amazed it worked so easily for me. Here's what it looks like:

1) Summarize what I think my project currently does

2) Summarize what I think it should do

3) Give a couple of hints about how to do it

4) Watch it iterate a write-compile-test loop until it thinks it's ready

I haven't added any files or instructions anywhere, I just do that loop above. I know of people who put their Claude in YOLO mode on multiple sessions, but for the moment I'm just sitting there watching it.

Example:

"So at the moment, we're connecting to a websocket and subscribing to data, and it works fine, all the parsing tests are working, all good. But I want to connect over multiple sockets and just take whichever one receives the message first, and discard subsequent copies. Maybe you need a module that remembers what sequence number it has seen?"

Claude will then praise my insightful guidance and start making edits.

At some point, it will do something silly, and I will say:

"Why are you doing this with a bunch of Arc things? Let's share state by sharing messages!"

Claude will then apologize profusely and give reasons why I'm so wise, and then build the module in an async way.

I just keep an eye on what it tries, and it's completely changed how I code. For instance, I don't need to be fully concentrated anymore. I can be sitting in a meeting while I tell Claude what to do. Or I can be close to falling asleep, but still be productive.

Re: A staff engineer's journey with Claude Code

#24

Post by conception »

MikeTheGreat wrote: post #14
swframe2 wrote: post #3

Preventing garbage just requires that you take into account the cognitive limits of the agent. For example ... 1) Don't ask for large / complex change. Ask for a plan but ask it to implement the plan in small steps and ask the model to test each step before starting the next. 2) For really complex steps, ask the model to write code to visualize the problem and solution. 3) If the model fails on a given step, ask it t…

Genuine question: What do you mean by " ask it to implement the plan in small steps"? One option is to write "Please implement this change in small steps?" more-or-less exactly Another option is to figure out the steps and then ask it "Please figure this out in small steps. The first step is to add code to the parser so that it handles the first new XML element I'm interested in, please do this by making the change X…

I tell it to generate a todo.md file with hyper atomic todos each requiring 20 loc or less. Then have it go through that. If the change is too big, generate phases (5-25) and then do the todos for each phase. That plus some sort of reference docs/high level plan keeps it going along all right.

Re: A staff engineer's journey with Claude Code

#26

Post by MangoCoffee »

swframe2 wrote: post #3

Preventing garbage just requires that you take into account the cognitive limits of the agent. For example ... 1) Don't ask for large / complex change. Ask for a plan but ask it to implement the plan in small steps and ask the model to test each step before starting the next. 2) For really complex steps, ask the model to write code to visualize the problem and solution. 3) If the model fails on a given step, ask it t…

I've been vibe coding a couple of personal projects. I've found that test-driven development fits very well with vibe coding, and it's just as you said break up the problem into small, testable chunks, get the AI to write unit tests first, and then implement the actual code

Re: A staff engineer's journey with Claude Code

#27

Post by groby_b »

awesome_dude wrote: post #19
axus wrote: post #18

I like his point about more objectivity and zero ego. You don't have to worry about hurting an AI's feelings or your own when you throw away code.

But I still find myself needing (strongly) to let Claude know when it's made a breakthrough that would have been hard work on my own.

Curious: Do you also laud your compiler for particularly good optimizations?

Re: A staff engineer's journey with Claude Code

#28

Post by rmonvfer »

swframe2 wrote: post #3

Preventing garbage just requires that you take into account the cognitive limits of the agent. For example ... 1) Don't ask for large / complex change. Ask for a plan but ask it to implement the plan in small steps and ask the model to test each step before starting the next. 2) For really complex steps, ask the model to write code to visualize the problem and solution. 3) If the model fails on a given step, ask it t…

I’d like to add: keep some kind of development documentation where you describe in detail the patterns and architecture of your application and it’s components.

I’ve seen incredible improvements just by doing this and using precise prompting to get Claude to implement full services by itself, tests included. Of course it requires manual correction later but just telling Claude to check the development documentation before starting work on a feature prevents most hallucinations (that and telling it to use the Context7 MCP for external documentation), at least in my experience.

The downside to this is that 30% of your context window will be filled with documentation but hey, at least it won’t hallucinate API methods or completely forget that it shouldn’t reimplement something.

Just my 2 cents.

Re: A staff engineer's journey with Claude Code

#29

Post by Barbing »

noosphr wrote: post #9
jason_zig wrote: post #4

Earlier quoted context omitted.

I've seen people post this same advice and I agree with you that it works but you would think they would absorb this common strategy and integrate it as part of the underlying product at this point...

The people who build the models don't understand how to use the models. It's like asking people who design CPUs to build data-centers. I've interviewed with three tier one AI labs and _no-one_ I talked to had any idea where the business value of their models came in. Meanwhile Chinese labs are releasing open source models that do what you need. At this point I've build local agentic tools that are better than anythin…

Very interesting. And plausible.

Doesn't specifically seem to jive with the claim Anthropic made where they were worried about Claude Code being their secret sauce, leaving them unsure whether to publicly release it. (I know some skeptical about that claim.)

Re: A staff engineer's journey with Claude Code

#30

Post by CuriouslyC »

adastra22 wrote: post #21
swframe2 wrote: post #3

Preventing garbage just requires that you take into account the cognitive limits of the agent. For example ... 1) Don't ask for large / complex change. Ask for a plan but ask it to implement the plan in small steps and ask the model to test each step before starting the next. 2) For really complex steps, ask the model to write code to visualize the problem and solution. 3) If the model fails on a given step, ask it t…

This is why the jobs market for new grads and early career folks has dried up. A seasoned developer knows that this is how you manage work in general, and just treats the AI like they would a junior developer—and gets good results.

Why bother handing stuff to a junior when an agent will do it faster while asking fewer questions, and even if the first draft code isn't amazing, you can just quality gate with an LLM reviewer that has been instructed to be brutal and do a manual pass when the code gets by the LLM reviewer.
Post reply on HN