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

251–260 of 410 posts

Re: A staff engineer's journey with Claude Code

#251

Post by sirpalee »

rhubarbtree wrote: post #212

Does anyone have a link to a video that uses Claude Code to produce clean robust code that solves a non trivial problem (ie not tic tac toe or a landing page) more quickly than a human programmer can write? I don’t want a “demo”, I want a livestream from an independent programmer unaffiliated with any AI company and thus not incentivised to hype. I want the code to have subsequently been deployed in production and de…

I had success with it on a large, established project when using it for refactoring, moving around functions, implementing simple things and writing documentation. It failed when implementing complex new features and horribly went off the rails when trying to debug issues. Almost all its recommendations were wrong, and it kept trying to change things that certainly weren't the problem.

Re: A staff engineer's journey with Claude Code

#252

Post by thecupisblue »

coffeeri wrote: post #238
rhubarbtree wrote: post #212

Does anyone have a link to a video that uses Claude Code to produce clean robust code that solves a non trivial problem (ie not tic tac toe or a landing page) more quickly than a human programmer can write? I don’t want a “demo”, I want a livestream from an independent programmer unaffiliated with any AI company and thus not incentivised to hype. I want the code to have subsequently been deployed in production and de…

This video [0] is relevant, though it actually supports your point - it shows Claude Code struggling with non-trivial tasks and needing significant hand-holding. I suspect videos meeting your criteria are rare because most AI coding demos either cherry-pick simple problems or skip the messy reality of maintaining real codebases. [0] https://www.youtube.com/watch?v=EL7Au1tzNxE

Great video! Even more, shows a few things - how good it is with such a niche language but also exposes some direct flaws.

First off, Rust represents quite a small part of the training dataset (last I checked it was under 1% of code dataset) in most public sets, so it's got waaay less training then other languages like TS or Java. You added 2 solid features, backed with tests and documentation and nice commit messages. 80% of devs would not deliver this in 2.5 hours.

Second, there was a lot of time/token waste messing around with git and git messages. Few tips I noticed that could help you in the workflow:

#1: Add a subagent for git that knows your style, so you don't poison direct claude context and spend less tokens/time fighting it.

#2: Claude has hooks, if your favorite language has a formatter like rust fmt, just use hooks to run rust fmt and similar.

#3: Limit what they test, as most LLM models tend to write overeager tests, including testing if "the field you set as null is null", wasting tokens.

#5: Saying "max 50 characters title" doesn't really mean anything to the LLM. They have no inherent ability to count, so you are relying on probability, which is quite low since your context is quite filled at this point. If they want to count the line length, they also have to use external tools. This is an inherent LLM design issue and discussing it with an LLM doesn't get you anywhere really.

Re: A staff engineer's journey with Claude Code

#254

Post by sunnyam »

boesboes wrote: post #237
rhubarbtree wrote: post #212

Does anyone have a link to a video that uses Claude Code to produce clean robust code that solves a non trivial problem (ie not tic tac toe or a landing page) more quickly than a human programmer can write? I don’t want a “demo”, I want a livestream from an independent programmer unaffiliated with any AI company and thus not incentivised to hype. I want the code to have subsequently been deployed in production and de…

I've been using it to do all my work for the last month or two and have decided it's not worth it. I haven't made any recordings or anything, so this is purely my subjective experience: it's ok at greenfield stuff with some hand-holding to do things properly all the time. It knows the framework well, but won't try to use it correctly and go off on weird detours to 'debug' things that fail because of it. But on a bigg…

I have the same opinion, but my worry with this attitude is that it's going to hold me back in the long run.

A common thread in articles about developers using AI is that they're not impressed at first but then write more precise instructions and provide context in a more intuitive manner for the AI to read and that's the point at which they start to see results.

Would these principles not apply to regular developers as well? I suspect that most of my disappointment with these tools is that I haven't spend enough time learning how to use them correctly.

With Claude Code you can tell it what it did wrong. It's a bit hit-or-miss as to whether it will take your comments on board (or take them too literally) but I do think it's too powerful a tool to just ignore.

I don't want someone to just come and eat my cake because they've figured out how to make themselves productive with it.

Re: A staff engineer's journey with Claude Code

#255

Post by hvb2 »

M4v3R wrote: post #243
rhubarbtree wrote: post #212

Does anyone have a link to a video that uses Claude Code to produce clean robust code that solves a non trivial problem (ie not tic tac toe or a landing page) more quickly than a human programmer can write? I don’t want a “demo”, I want a livestream from an independent programmer unaffiliated with any AI company and thus not incentivised to hype. I want the code to have subsequently been deployed in production and de…

I've live streamed how I've built a tower defense game over a span of a week entirely using AI. I've also written down all the prompts were used to create this game, you can read about it here: https://news.ycombinator.com/item?id=44463967 Mind you I've never wrote a non-trivial game before in my life. It would take me weeks to do this on my own without any AI assistance. Right now I'm working on a 3d world map edito…

> What matters is I manage to finish projects that I would not otherwise if not for the AI coding tools, so having them is a huge win for me.

I think the problem is in your definition of finishing a project.

Can you support said code, can you extend it, are you able to figure out where bugs are when they show up? In a professional setting, the answer to all of those should likely be yes. That's what production code is.

Re: A staff engineer's journey with Claude Code

#256

Post by wooque »

rhubarbtree wrote: post #212

Does anyone have a link to a video that uses Claude Code to produce clean robust code that solves a non trivial problem (ie not tic tac toe or a landing page) more quickly than a human programmer can write? I don’t want a “demo”, I want a livestream from an independent programmer unaffiliated with any AI company and thus not incentivised to hype. I want the code to have subsequently been deployed in production and de…

You got it wrong, the purpose of this blog post is not marketing Claude Code, but marketing their company. Writing about AI just happens to get more eyeballs.

Re: A staff engineer's journey with Claude Code

#257

Post by infamousclyde »

rhubarbtree wrote: post #212

Does anyone have a link to a video that uses Claude Code to produce clean robust code that solves a non trivial problem (ie not tic tac toe or a landing page) more quickly than a human programmer can write? I don’t want a “demo”, I want a livestream from an independent programmer unaffiliated with any AI company and thus not incentivised to hype. I want the code to have subsequently been deployed in production and de…

Jon Gjengset (of MIT Missing Semester, Rust for Rustsceans, etc) shared a stream doing complex changes of increasing complexity to a geospatial math library in Rust. He’s an excellent engineer, and was able to pick apart AI-suggested changes liberally. The caveat is that the video is a bit long, but segmented nicely.

I think he had a positive experience overall, but it was clear throughout the stream that he was not yielding control to a pure-agent workflow soon.

https://youtu.be/eZ7DVHAK8hw?si=vWW4kz2qiRRceNMQ

Re: A staff engineer's journey with Claude Code

#258

Post by ffsm8 »

hvb2 wrote: post #255
M4v3R wrote: post #243

Earlier quoted context omitted.

I've live streamed how I've built a tower defense game over a span of a week entirely using AI. I've also written down all the prompts were used to create this game, you can read about it here: https://news.ycombinator.com/item?id=44463967 Mind you I've never wrote a non-trivial game before in my life. It would take me weeks to do this on my own without any AI assistance. Right now I'm working on a 3d world map edito…

> What matters is I manage to finish projects that I would not otherwise if not for the AI coding tools, so having them is a huge win for me. I think the problem is in your definition of finishing a project. Can you support said code, can you extend it, are you able to figure out where bugs are when they show up? In a professional setting, the answer to all of those should likely be yes. That's what production code i…

I disagree with your sentiment.

The difference isn't what's finishing a project is, is the dissonance between what M4v3R and rhubarbtree understand when talking about "nontrivial production" software.

When you're working in enterprise, you usually have multiple stakeholders each defining sometimes even conflicting requirements to behavior of your software. And you're required to adhere to these requirements stringently.

That's an environment that's inherently a bad fit for vibe coding.

It can still be used there, too, but you will not get a 2-3x speed up, because the LLM will always introduce minor behavioral changes - which aren't important in M4v3R scenario, but a complete deal breaket for rhubarbtree.

From my own experience, I don't get a speed up at all via CoPilot agentic mode (Claude code is banned at my workplace). But I have had a significant boost in productivity in projects that don't need to adhere to any specific spec - namely projects I do an my own time (with Claude code right now).

I still use Copilot agentic mode though. While I haven't timed myself, I don't think I'm faster with it whatsoever. It's just less mentally involved in a lot of scenarios, so it's less exhausting - leaving more energy for side projects .

Re: A staff engineer's journey with Claude Code

#259

Post by sanitycheck »

mindwok wrote: post #171
ale wrote: post #45

It’s about time these types of articles actually include the types of tasks being “orchestrated” (as the author writes) that aren’t just plain refactoring chores or React boilerplate. Sanity has quite a backlog of long-requested features and the message here is that these agents are supposedly parallelizing a lot of the work. What kind of staff engineer has “80% of their code” written by a “junior developer who doesn…

IMO “junior developer who doesn't learn“ is not quite right. Claude is more like an senior, highly academic engineer who has read all the literature but hasn't ever written any code. Amazing encyclopaedic knowledge, zero taste. I've been building commercial codebases with Claude for the last few months and almost all of my input is on taste and what defines success. The code itself is basically disposable.

Eh, Claude is like a magical spaniel that can read and write very quickly, with early-stage alzheimers, on amphetamines.

Yes it knows a lot and can regurgitate things and create plausible code (if I have it run builds and fix errors every time it changes a file - which of course eats tokens) but having absolutely no understanding of how time or space works leads to 90% of its great ideas being nonsensical for UI tasks. Everything is needing very careful guidance and supervision otherwise it decides to do something different instead. For back end stuff, maybe it's better.

I'm on the fence regarding overall utility but $20/month could almost be worth it for a tool that can add a ton of debug logging in seconds, some months.

Re: A staff engineer's journey with Claude Code

#260

Post by nosianu »

Kiro wrote: post #231
rhubarbtree wrote: post #212

Does anyone have a link to a video that uses Claude Code to produce clean robust code that solves a non trivial problem (ie not tic tac toe or a landing page) more quickly than a human programmer can write? I don’t want a “demo”, I want a livestream from an independent programmer unaffiliated with any AI company and thus not incentivised to hype. I want the code to have subsequently been deployed in production and de…

Very few people want to record themselves doing stuff or have an incentive to convince anyone except for winning internet arguments.

> Very few people .... have an incentive to convince anyone

We are already only talking about the subset the writes AI blog posts, not about all of humanity.

Post reply on HN