upvote
>the AI codes better than you.

A blocker exists here. I've personally talked to VPs and managers that are simply behind and refuse to acknowledge this is true. If you're applying to work and state this in an interview, they will immediately decide they need someone else, someone who can code better than AI every damn day. You cannot talk your way into a job after you say this and the decision-maker disagrees. Some things simply cannot be helped.

reply
You don't want a job like this. You want a job where management lives in reality and gives you the tools you need to succeed.
reply
A job where you are valued as being better than AI seems better than one where the boss knows you are just the minimum wage yes-man giving encouragement to Claude to finish the work.
reply
Well this is where reality bites. In the abstract what you said sounds great. Concretely, if your boss told you that you can dig faster with a spade than a backhoe, it suddenly doesn't sound so great. I suppose it's possible that there are minimum wage SWE jobs but that's just not my experience.

/goal exists for encouragement loops. If someone just needs an encouragement loop they are not hiring right now. The job is much more than coding and always has been much more than that.

reply
I think we agree, but I maybe was not clear with my statement. My manager thinks I am better than AI, but implicit in that is that I will use AI as a tool when appropriate. He does not give any indication he thinks AI would do my job better than me (I think at this point he is way more worried about his own future).

But I'm very senior and so I haven't been predominantly a coder for a long time now. I keep things on the rails and headed the right direction, grease the rails, cajole teammates and stakeholders, etc.

reply
Another example: if you are already tutoring math, how can you use AI to automate all back-office/non-math-tutoring parts of the job?

examples: Managing your calendar, responding to emails from your students (with human in the loop review), sending invoices, following up on unpaid invoices, writing/updating your website, creating/managing your yelp/google business accounts, emailing past students for referrals, etc etc.

All of these things are very interesting from an "AI Orchestration, AI Software Factory, MCP to integrate several source systems, etc" perspective that can make your 2 years of unemployment really shine.

And the good news - you don't have to lie and say "I was building with AI the whole time", you can say "I made a career change to math tutoring for X reason, learned everything needed to run the business manually, then I recently started automating all of it like a Forward Deployed Engineer would do, and here's where I ended up so far".

reply
> Coding is not a job anymore. It has evolved.

You're narrowly defining coding as "writing lines of a programming language by hand." But coding, or programming as it is better called, is about creating software. It is about creating a program. A task for a computer to do. And whether it is done using English, Fortran, or assembly it is absolutely still a job.

So while you're right to point out that C# is not as valuable a skillset as it was before and people are concentrating more on higher level business and architecture requirements, the end of your post is outrage bait.

reply
The fact that he claims to deliver stuff at "breakneck speeds" is a tell he hasn't done anything of value. Or he works on Teams.
reply
Comments like theirs make me want to kms
reply
Agree, language, frameworks and tools are just a word in a prompt now, you can add that as your pre-AI experience but it's not relevant today. You should be showing how you know how to bend frontier models and tools to your will via, agents, MCP, API's etc. Taking a task that used to take weeks to hours.

But even this is short-lived, and the writing has been on the wall for years. Hopefully you invested hard in companies that will grow as AI diffuses into the economy so your assets can throw off more income than you earned in the gravy train years of SWE. If not your livelihood is going to drop, even starting a business is unlikely to succeed as everyone else has AI to help them so you coding skills give limited advantage there.

reply
> the business domains

This is represented by code too though. Or at least some form of written documentation.

Sure, you won't use anything lower level than typescript for it normally, but for complex relations and constraints, code built on your own abstractions (and with types) is still easier to reason about and make robust than some non-technical person vibecoding in English without any good workflows or collection of skills (that they actually know how to use).

Agreed, "coding" is not a job. But software engineering 100% is, design 100% is, and devops 100% is.

reply
Well said. Coding was always a tool to make things happen. Now that writing code is automated, people should focus on the parts where they can still make a difference.
reply
This is terrible advice. Knowing the ins and outs of specific languages enables you to use an LLM with them. What does “codes better” even mean?

> check out the /wayfinder skill from Matt

How is the OP supposed to know what this means? Seems like borderline psychosis to think this is just known. Personal experience has been people who use these tools are kinda low skill, too.

reply
I copied that phrase into ChatGPT. It gave me the link, described the process, told me how it is applicable to one of my projects but too heavyweight for another. And concluded that it resembles classical software engineering more than vide coding. Hmm.
reply
If you Google "/wayfinder by Matt" the first result is https://www.aihero.dev/skills-wayfinder but I guess expecting the bare minimum legwork is psychosis now
reply
That's what I did. I had not heard of it. And, I have a large spec that may be fruitful to throw at it to break it up into a better more manageable time line and tasks.

Looks interesting.

reply
> The AI codes better than you

Where is this mythical AI? I have not found it yet. I'm not doing C# with AI, but whenever I make AI code something (and yes this is with frontier models), both the quality and the quantity are visibly and significantly worse than a human engineer's. The code is frequently brittle and abstraction-breaking, and the tests are simply garbage. And I know this isn't specific to me because I also see this - consistently - with what other engineers send me to review too.

I would more readily buy "code quality doesn't matter now that we have AI" than "AI codes better than you."

reply
Same experience, AI gets things sort of working fast, but the code quality is generally quite bad.
reply
This is a skill issue. If you know best practices for software engineering, you can make your AI follow them. If you tell a model to reimplement Google Earth make no mistakes, yeah, it's not going to write high quality code. If you have a workflow for generating spec documents that you actually review before telling your agent to implement them, you'll get high quality, well specified code out.
reply
would you be down for a head to head on a live stream, where you code without AI against someone using AI and have the results judged by humans? small prize to the winner - I can’t find many participants for the non-AI side, one to two hours long
reply
Speed is only one factor, quality should matter.
reply
How can you measure quality without time boxing, and therefore measuring speed?
reply
Quality is orthogonal to time taken although likely correlates positively.

Michelangelo's David took years to complete, many studies and iterations. The patrons that paid for it wanted the perfection such an artist could provide - that was not something to be rushed.

Software that is ergonomic, intuitive, fast, and maintainable takes time. To use the analogy - chipping away at granite is a slow and tedious process, mistakes in planning or approach are difficult if not impossible to recover from.

Taking time to do it right sometimes is the only way to achieve quality.

reply
The contest doesn't make sense. 1 hour with access to AI is clearly going have an advantage over 1 hour without - nobody claimed otherwise.

The problem is that, in the real world, people are neither competing for maximizing the code quality score, nor are they spending remotely the same amount of time with vs. without AI. They're spending a small fraction of the time with AI than they otherwise would have, and their goal is to ship the product. Hence why they commit bad code.

To put it bluntly, I've literally seen engineers who used to send excellent pre-AI code suddenly start sending trash for review when they use AI. And, when I push back and point out the problems and they actually put in the effort to improve it, their code quality improves to (or beyond) what they (or I, for that matter) would've written without AI. I don't know how you could get a more fair experiment than this.

So, to answer your question directly, no. (No for other reasons too, but this is the only relevant one.)

reply
I'm convinced these comments are written by people who tried AI 2 years ago and have never used it since. Because there's no way you can be using SOTA models and have this opinion.

I simply don't believe you.

reply
Well I guess we're in the same boat then because this is what I have been seeing up to 2 days ago so I simply cannot believe what I'm reading either.
reply
I use Opus 5.5 and Fable 5.2 at work, in Claude Code set to high or max reasoning. It smokes me in terms of code quality for anything I do at work. But at work I have to use tech that the company picked, work in a codebase that others contribute to, and build out features that stakeholders are asking for.

For personal projects where I have a deep understanding of what I want to build and genuinely care about the quality, even Fable can't touch me. Of course, AI would get a working system done 50x faster, but personal projects aren't about speed for me.

reply
it's obviously because everybody on HN is working on some cutting edge stuff that even the most SOTA model can't do effectively /s
reply
I find the SOTA models are far better at developing new cutting-edge stuff than boring updates to existing code, so this sarcasm doesn't even make sense to me.
reply
The quality of the code is irrelevant, the only two metrics that matter are a) security and b) performance. If the AI produces code that is secure enough (aka clearly passes security checks/fuzzes/linters/LLM reads) and perform well enough in general and targeted benchmarks then it's good enough. The actual code style, abstractions are by-and-large irrelevant today. An LLM could even insert inline ASM into Rust, and if it performs well, well, it performs well.
reply
> the only two metrics that matter are a) security and b) performance

You don't care at all about (c) maintainability or fragility?

One of the biggest problems I see is LLMs duplicate sources of truth they have no business duplicating... all the time. E.g. if an internal constant in some code is X then they just hard-code X in fifty other places far away and code based on that. And it's not just constants, they often duplicate entire pieces of logic too. So everything is fine and dandy when the code is fresh, but the next time the original is updated, the rest of the code breaks, because nothing forces this update to update too. Of course the garbage tests they produce never test such hidden assumptions either.

And this is before I get to other things, like (d) test costs and (e) the limits of test matrices. That "inline assembly" you just inserted now requires you to test everything on older CPUs that may lack newer instructions. And on other architectures too. That's a huge cost and this kind of thing exponentially grows your test matrix. Or, more likely, you end up with a breakage on some CPU because you didn't care.

reply
> The quality of the code is irrelevant

If you have unlimited tokens i guess this is pretty much true these days. I've noticed opus write write and remove the same piece of code in a session. They prefer to write a lot, a lot of defensive checks in python etc.

At my company I have around 4000 GitHub copilot credits per month. I cannot afford to just let the LLM finish a whole project by itself using just spec documents. I did that once, ran out of tokens and then found out i couldn't understand what was going on.

This will probably change in the future but right now I need to work with what's available to me

reply
and what about stability, quality, convenience metrics? You think they do not matter?
reply
Quality and convenience only matters insofar as it affects LLMs ability to iterate on the codebase in the future. Stability is a matter of correctness and included in performance imo, ie is it performing the functions correctly in reasonable time
reply
Ah, this is why so much software is incredibly buggy these days.
reply
>Where is this mythical AI?

I'll be blunt: Claude Code using Fable High. But, it's not the only one, you can't hardly swing a stick without hitting an AI that is great at coding. If by coding you mean "produces working software". If your definition is "produces code that doesn't offend my sensibilities", then that's a different matter, and also likely something that a human expert developer likely couldn't accomplish either. ;-)

reply
Well we seem to live in two different universes then, because that's not my experience with Fable High or anything else: https://news.ycombinator.com/item?id=50589372
reply
I'm having a hard time understanding what point you are making with that link. It seems to boil down to: AI is making better PRs than I or the contributor could have made. But the caveat is that the contributors aren't getting it right the first time, but you don't go into the details about why.

Or are you saying that the "mythical AI" is mythical because people are using it without putting in any effort, therefore producing poor results? If that's the core of your argument, it sounds like the mythical AI could be out there and you'd not know it because those around you are, to coin a phrase, using it wrong.

I'm kind of struck though that your experience seems to all be second hand (people are using AI and submitting bad code); do you not have any first-hand experience with using claude code and fable?

reply
> I'm having a hard time understanding what point you are making with that link

I'm saying that most code I see AI generate is frequently bad, in ways that nobody disputes, not even the engineers who sent me the code to review - they themselves previously always avoided those exact problems when they used to write the code by hand.

> Or are you saying that the "mythical AI" is mythical because people are using it without putting in any effort, therefore producing poor results?

This also happens.

> If that's the core of your argument, it sounds like the mythical AI could be out there and you'd not know it because those around you are, to coin a phrase, using it wrong.

If you hired a brilliant math PhD who always wrote bad code and you had to keep reviewing every single thing they did and getting them to rewrite their code until it met the quality bar every time, and they never learned from their mistakes from one project to the next so this was a persistent issue, but they were so fast that even micromanaging them let you achieve a better outcome faster than you would have alone, what would you call that? That's what AI feels like.

> do you not have any first-hand experience with using claude code and fable?

I certainly do. But for anything meaty I frequently end up rejecting most of the code they produce before they get to other reviewers, and even then, I frequently have to do more work on them by hand. In some cases I've quite literally had to restart a project half a dozen times from scratch - with different approaches each time - in order to reach results that I'm not disgusted with.

Note that I'm not claiming to be an expert in using AI by any stretch - I'm still very much learning - but that I'm not really seeing these mythical results from others either.

reply
you have to tell it not to do that stuff, also by telling it what to do instead
reply
Orchestrating workflows is not a real skill and will be automated away soon, after the discovery phase where we figure out what works best for llms.
reply
Maybe not orchestrating, but deciding what parts of the app should be verified during the agent loop is for sure a skill.

And shrinking the context bloat while maintaining intent and debugability is also something you'll have to get better at.

Even the experience of using a Terra-type model vs using Opus requires a different workflow because one can just go on longer and make reasonable decisions most of the time.

reply
Again this is simply a software issue that is easily generalizable, and the constraint you mention comes from having to use a weaker model. Given how they are progressing and decreasing in API prices, it is likely to be solved soon.

Kind of like driving a car doesn't require you to know how to do basic fixes on the road anymore.

reply
> the AI codes better than you

Hahaha I spent hours yesterday attempting to build a solid multi tenant api with Claude code in .net. I refined it for a couple of hours, came up with a plan (wrote it to markdown files), used Opus and Sonnet to write it and used code review. It was a disaster. Loads of code but loads of terrible decisions, bugs and suboptimal data design.

14 day session expiry, active org ID in the cookie, single tables with nullable keys instead of splitting it up...I can't recall the other issues and bugs as every review for about 4 reviews in a row threw up 1-3 problems (and I don't have my laptop to view the git log)

It often just straight up ignored any hard parts of the plan.

The tests were brittle and slow despite running the tests for every change it didnt stop and fix it. When challenged it was sassy like "I think 15s isn't too bad"

Something 'super intelligent' trained on billions of lines of code should not need every single aspect of every single module explained. A secure Auth and multi tenant system is hardly novel

It basically generates so much code it hopes you won't notice all the issues bugs and footguns. Just like bullshitters like who don't shut up hoping you can't keep track of their bs

reply
How long should that take an independent human to write with the accuracy you need?
reply
Yeah that's not really a gotcha because it's not just time but accuracy and security. It's following established practices, OWASP/NIST recommendations. It's not causing painful data issues. Not making the test suite a ball and chain almost off the bat etc

Again, it'll happily produce lots of code quickly to pacify people who want to stick their head in the sand and pretend lots of code which looks like it ticks some boxes must be correct. It seems to excel at that (I've been using Claude for various other small projects in different languages btw, this isn't my only experience with it)

"Fast, cheap, good - pick two" seems still to hold

reply
> some harsh feedback

This is all obviously false. Thinking that "coding is not a job anymore" is as delusional as being anti-AI.

But I guess this is what the industry wants to hear: sycophancy towards AI and hatred towards human programmers.

reply
>Thinking that "coding is not a job anymore" is as delusional [...]

The best coder in the company I work for just said "I haven't opened an IDE in a couple months". He was a die-hard JetBrains guy, now he's talking about letting his subscription expire. I'm in SRE, and so I'm not a day-to-day coder, but I've been coding for 40+ years and it's really obvious in SRE.

I can imagine some corner cases where coding is still a job, but for the general case I think there's a good case to be made today, and can absolutely see we are on a trajectory where it's only getting more true.

If anything, coding may be something we do to FURTHER our job.

reply
Yep, harsh but apt.
reply