We write code by hand | Hacker News Reader

We write code by hand

handcraftedcode.org

42 pointsbatmang71 comments

I feel like I have gone from 100% hand written code -> 0% hand written -> 50% and climbing.

I think there are things LLMs are really good at coding up and experimenting with but I think there's still a need to understand the btoader context of your software so you need to dive in at some point.

It's very important to understand what's happening, sure. Blindly shipping AI-produced code without understanding is just irresponsible. But how does writing code helps with that? Reading helps a lot, but writing just slows things down in my experience so far.

You can't learn what good code is by reading it, you have to write code to learn what good code looks like

True, but having written enough (from many years of doing it full time), the returns from writing more are diminishing.

But yes I agree that as part of learning effort it's important to write, too.

I'll push back against this slightly: learning what good code is requires debugging bad code, whether by way of a symbolic debugger or judicious use of printf() or breaking an algorithm down into its steps so that you can figure out its asymptotic behaviors.

I harped on checking someone's ability to debug code back when I interviewed candidates, in part because I found that people who could debug effectively could also sniff out bad code. It was something of a bonus that I also happened to select for candidates who could explain their debugging strategies and what that meant for the software they were working on.

I don't really draw much distinction between debugging and writing code. You really can't do one very effectively without the other imo

But you're right. The better indication of capability is debugging skill than writing skill.

It's kind of the same distinction between backups and restores, wherein the restores are what's important, but they're impossible without the backups to restore from.

One of the first things I do in any new-to-me project is set breakpoints in the integration tests and start stepping through the code so that I can establish some mental context. However, if there aren't integration tests, I begin writing them so that I can walk around with my debugger.

I compare it to math, you can't learn it by reading it you need exercises. Same thing, reading is not sufficient. Just my experience.

Yeah if it's done as part of learning effort, then I def agree.

Hard agree on this. I feel like it's pretty normal, but I struggle with retaining information by just reading. I really need the manual typing step to cement the knowledge.

LLMs will make tradeoffs without consulting you. This is sometimes a feature and often a bug. Many times a decision made up front can have non obvious implications down the line. This is one of the biggest values I bring to be table as a more experienced swe.

I think if I fully understand everything the agent is going to do and trivially verify the output quality, LLMs are a win. For one off prototypes, the same is true. For software which is unique, complex, and performing a task which has not been fully specified I find myself needing to drop into my editor more and more.

Recently I've been moved to a research focused team and I really need to have proof that something is happening. I've had Claude Opus 5 gaslight me by telling me it did something and when I read the code it obviously did not. This happens more and more with my tasks that are kind of complex.

I've found a lot of the claims by AI people have been 6mo - 2 years "ahead" of my experience. I think we are now in an era where harness engineering is highly valuable (making a test, looping an agent, manually annealing with new ideas) but the claims that no one writes code manually seems like it may not be fully there for all code.

I had a similar experience but I'm closer to 95-99% hand written again now. I still use AI for reviewing and prototyping, but it feels like writing the code myself is the best way to actually know what's going on.

I like to write the edges between modules, because that’s easier than prompting. Think interfaces, or public func signatures.

The language here means that signers are affirming that they continue to write at least some code by hand, not that they never use AI tools for coding. I'm not sure that's enough virtue to deserve its own signal.

edit to add: the irony of the grammatical error in this presumably hand-written sentence inside the "Who's behind this?" link:

> Jared’s web studio Whitefusion is currently partnering with quality businesses & organizations which believe in staying away from costly hype cycles and the value of the human touch.

Self-plug but Handmade Cities [0] is grassroots organizing that prefers not to spend its time with stuff like this.

Join a local meetup!

[0] https://handmadecities.com

I began programming as means to express myself artistically, but over time, it became more of a way to value myself instead. AI is helping me get back to the former, and with that I have peace.

I like this sentiment. It's the same I share as well.

To me, literally writing code by hand instead of letting AI do it feels like digging a foundation pit with a shovel instead of letting an excavator do it. Sure it can be done but much slower and the benefits are questionable. But curious if you have reasons to believe otherwise?

If your work is analogous to digging a foundation pit or something like framing up homes in a development, that makes a lot of sense. But that's not the only kind of work many of us do, and analogies to things like finish carpentry, functional sculpture, or fine furniture are more applicable. Hand craft doesn't always make a difference, but there's plenty of places where it does.

Plus, for a lot of people with fluency and experience, it's simply more fun/natural and not meaningfully less productive.

It's not the same example, the excavator is driven by a human being, it's a tool (in coding world the same as using an high level programming language instead of writing bits into the computer memory).

The LLM *may* be used as a mere tool, that is something you ask question time to time, but shouldn't be the thing that makes the work that you should do. Because if we arrive at this point (hopefully never) at that point the CEO of the tech companies may as well say, why we still need developers? Let's fire all of them.

Fortunately they did not, because there is STILL VALUE in writing code by hand, understanding what it happens, what the code will do, etc. I hope this will not be a lost skill, or well, if it does good for me, because the same as nowadays things being able to repair electronic devices makes a ton of money (I've considered opening a repair shop) because there is no one still doing it, will do programmers that well, know how to program.

Yes, LLM is a tool. And CEOs of the tech companies are retaining developers not because there is value in writing code by hand, it's because developers can use that tool effectively and responsibly: they can tell bad code from good code, they design the solution, let AI do the hard boring work of coding it out, then these developers check its output, test it, also they can be creative and e.g. ask LLM to review LLM's output, etc etc.

But the need to manually type stuff out arises very infrequently in my experience. It happens but it's an exception, not the rule.

Every excavator crew I've seen has a shovel on the truck. They sometimes pull it out of the truck and use it. Most of the time though you can get the excavator there and get the job done faster than any hand work.

The most common use for a shovel seems to be "this is as close as I dare get to the wire we know is around here - scrape the dirt away until you find it and then I'll dig the rest out."

Yeah a small amount of manually-written code still makes sense, when it's faster to write it than to tell LLM to write it.

In particular comments. I'm not a fan of the comment style LLM tends to write, so I often just write them manually.

And also coming up with the right architecture is also often manual job. Design how to implement things, and let LLM implement it.

Genuine question: do you feel this disparity despite being diligent and checking the LLM's output, or is that not factored in?

It's also been my experience that generating an entire app/portion of code can be much easier by typing out a relatively concise prompt, but this only applies if it's the kind of code I'm

1) free from having to verify afterwards,

2) not having to fight with the LLM to get right, or

3) is coming from someone else in the form of a PR that may/may not have been checked by them first.

Having to deal with these cases has sometimes been the equivalent of using a spoon to dig a foundation, at least in my experience.

Of course checking the LLM's output and testing it and iterating on it is factored in. And yeah that verification and testing and iteration where most of the time goes, and still it ends up being significantly faster than typing everything manually, and the end result quality is often higher I believe.

To avoid fighting with the LLM to get right, it's useful to first come up with the architecture manually, run it by the LLM first (without writing any code - just talking to it, asking it to find flaws in your architecture), and once both you and LLM are happy with the plan, fire it to execute. This step removes a lot of friction (but not all of it - later iterations are still almost always necessary)

This is a pretty good metaphor to explain why there is so much unmaintainable slop around these days.

if you write a prompt, and see the code and edit it by writing another prompt, you the human wrote that code. you may not have typed the keyboard for each character in the code. but the llm did not write it all for you. this is the same as a pair programmer, 2 authors

big agree here.

People dont seem to realize prompt to code is just another interface. Is the human review part that we should care about, did this code get reviewed and reiterated on? Is it fully AI generated and reviewed?

Those are the important tasks.

Whoever feels like a better programmer because they write all the terraform files by hand and not use an llm, dislike.

The difference between Vibe coding and AI-assisted coding. There is a gradation!

Comes across as this "I'm special", and kinda cringe...

"We Weave Fabric by Hand" - Weaver, England, 1812

This has the energy of a collective of archers proudly announcing that they still fire arrows in the age of gunpowder.

To each their own, though.

Perhaps more like snipers saying they don't need full-auto?

Its written like it's making a praiseworthy point, but mostly comes across as luddites.

Luddites were making a praiseworthy point.

in the 19th or 21st century?

Yeah

You use the word luddites as a pejorative, but don’t substantiate why that is the case. What’s the actual issue?

> Also with lower-case initial. disparaging. Opposed to new technology or ways of working; disliking technological change and innovation.

From the OED. The predominant use of "Luddite" has been disparaging for a long time.

You can debate whether it should be or not, but it's not reasonable to criticize conventional use.

Fair callout, I think more to the point of the parent poster and to what I intended to ask - why is it a bad thing if people want to code by hand.

There seems to be an unstated assumption here that "handcrafted" code is somehow _better_ or _more desirable_ than AI-written code, all else being equal. I would ask for evidence of that claim.

I've worked in corporate environments for nearly two decades, and I can assure you that I have seen a lot of very low-quality "handcrafted" code running in production. I don't really see "handcrafted" as a signifier of quality, even if I agree that SOME SPECIFIC handcrafted code is better than SOME SPECIFIC AI-written code (and vice versa).

Yeah, can't agree more, and my experience is pretty much the same as you describe.

It's true. The bad:good ratio of handcrafted code is about 7 or 8:1.

Sure, those 8 are often thrown away while that 1 often ends up moving mountains but so what?

For vibe coded code from what ive seen it's about 10:0.

We should clearly only write that from now on.

I've never written code by hand, I've always used a text editor on my laptop. I still design and develop software with my brain, and in conversation and collaboration with my team and stakeholders --- and these days frequently an LLM. Sometimes (often) I make edits by hand, but it has reached a point where generating code by prompt is so much faster that it allows me to allocate more time and brainpower to the actual bottlenecks.

I'm sure a lot of the signatories on this list feel the same.

I wonder if by signing they're saying they write zero code with AI. Because huge spectrum making the statement almost meaningless.

Aside: what are the icons beside people's names? Particularly the pentagram star one... Assuming something to do with the Fediverse, but not Mastodon? Yet seeing number of entries link to a Mastodon instance, but also some other sites like some kind of ActivityPub server etc... Wasn't the Fediverse logo supposed to be that three-asterisk thing?

Can you still use Claude/ChatGPT as a "better google" or would that not count as hand coding anymore? I guess if you type it out instead of copy and pasting it. That's not much different than the old way of using Stack Overflow but with the ability to ask more specific questions.

I love how the list is unironically filled with links to peoples bluesky's. X is a very AI-forward platform with people using it to build the future; Bluesky is presumably filled with old foggies who think 2k lines of code a sprint is impressive.

Oh, we're back to using lines of code as a productivity measure?

"build the future" meaning build a very specific form of future, which not everyone would rather be part of.

There is a niche market for hand-knit sweaters, and they do have some fervent followers, but most of us prefer the price and convenience of machine-made fabrics. Just like most of us do not write assembly by hand any more, most of the time.

I admire the will power of the people able to do this.

There's just no way I could go back to writing code by hand with the temptation of AI right there in front of me.

Good for you. Some people paint on canvases, too.

This is a list of people who have downgraded themselves from professional to hobbyist

and perhaps have less secure software, agents reviewing code catch more issues than humans, what are they shipping?

demonstrably insecure software

The amount of Mastodon icons on that page sadly also confirm my feeling that Mastodon in general is a very "Anti AI" bubble and it became a lot less fun to hang out there if you are even slightly interested in AI or working with it. Not even talking about being some AI hyper, just someone who sees benefits in some fields and my daily work life.

Oh sweet, free backlinks. Thanks for the SEO.

Backlink from a website with no authority doesn't really help you.

Do you think, code by hand worth it and why ?

Gross how many commenters are swarming to take pride in their ignorance and generation of deficient, bug-ridden code. Like actual cockroaches, there's 15 one-line comments to this effect in 10 minutes. LLMs are obviously immensely useful for certain types of tasks and prototyping but you have to be seriously incompetent if they're faster and better than you are for writing production code in any area where you have the slightest bit of expertise.

Organic artisanal code?

"We will not get off the saddle and refuse to buy a car!"

njoy mhy hnnd crfhted proze

It's wild to watch speciation in real time. (Quick, interpret that according to your chosen side.) I do not know who will "win", but the real-time divergence between the commentary here and, say, other "programmer news" sites is incredible to live through. This reminds me of mid-2020 days of COVID and the level of, eh, confidence going around about masks or lockdowns or vaccines.

I appreciate seeing this, but I would like to see more people talk about WHY handwriting code is still valuable. For me, it's not because I'm a purist, but when I tried drinking the coding agent kool-aid, I saw my productivity get worse rather than better. I've thought about sharing more on this, but I've seen a worrying trend where anyone who talks about how they're not seeing amazing productivity gains from AI gets told it's a skill issue and they're doing it wrong.

Anyway, I’ll briefly share why after trying full agentic coding in earnest, I’m back to writing almost all my code by hand (I even turned off AI edit predictions). I will not deny that a lot of AI code is quite good these days, but in my experience it’s nowhere near the point where I’m comfortable not reviewing it. So if I have to review it, then this means I need to read verbose more code than most human developers write and that is just not fun. But beyond just not being fun, as I got farther from writing the code myself, it felt like I was a worse reviewer. Obviously total skill atrophy doesn’t happen overnight, but I felt like even being away from coding for a few weeks or months, I was more and more “out of shape” and whenever I did dig into the code base, even if I had reviewed everything and guided the AI, it all felt really alien. And then there is the part about needing to read tons and tons of annoying AI generated prose and argue with it constantly. Ultimately after a few months of trying this on some personal projects, I realized that because I had been so disinvolved with writing the code, most of it needed to be reworked in pretty substantial ways. Once I started writing the code again, I realized my net ability to steer the code in the right direction was faster when I typed it out than by going through this tiring LLM prompt/review/re-prompt loop while my skills were subtly atrophying each day. Now that I’m back to writing almost all my code by hand, I sort of hold the opinion that hand-coding is an unreasonably effective away of developing software architecture intuition and steering a codebase in the right direction. Also, I realized typing code is not actually that slow - thinking is definitely the slower part and the more you offload your thinking and understanding to robot who has literally zero persistent memory or stake in your project, the less capable you’ll be of doing the thinking when you need it.

I still use AI for brainstorming, reviewing my code and one-off prototyping/benchmarking tasks, but usually even if it writes some decent code snippet in the chat for me, I’ll go and handwrite it myself because then I actually get to learn something! This also let’s me confirm that the AI suggested approach is correct (usually it’s pretty good, but sometimes it’s off). I know some people are saying you have to stop reviewing code if you want to be fast with AI - that’s true and maybe that’s what the future will look like. But if humans still need to be involved in understanding and guiding code, I think handwriting code is quite a good way to stay on top of things.

Love to see a blog about it!

Bruh of course this is coming out of Portland. That'll be the last city writing code by hand.

We Chop Trees by Hand

Purely handmade code will be the symbol of quality.

Yea every PR submitted before LLMs was top tier work. Humans are born with an innate instinct to code and build complex software. It's the great equalizer. There are not outliers in coding. It is with great pleasure we "LGTM" and rubber stamp contributions. Nothing ever went wrong until LLMs ruined it for everyone.

Reply on news.ycombinator.com