deterministic

The job of a software engineer has always been to turn a real-world idea or problem into a precise specification a computer can execute. That specification used to be assembler, then higher-level languages like C and Python. Now it might be detailed Markdown files that AI turns into code.

So our understanding moves from the C/Python level to the Markdown specification level. We still need to understand exactly what we're building and why.

The hard part was never translating the spec into machine code. The hard part is defining the right spec. That is still the job of the software engineer.

I am soooo tired of that quote. Especially because it’s BS. In 30+ years in the industry, I’ve never seen it happen. Not once. And I’ve never seen anyone demonstrate it convincingly on a real, large-scale commercial project either.

Jaw droppingly excellent work. Thanks!

I’m so tired of the “X written in Rust” posts on HN. Seriously, who cares what language you used? Especially nowadays, when LLMs are writing much of the code anyway.

What an excellent read. This is why I read HN!

This post doesn’t really add anything new. These points about AI and critical thinking have been made many times before.

The job of a software engineer has always been to turn a real-world idea or problem into a precise specification a computer can execute.

That specification used to be assembler, then higher-level languages like C and Python. Now it might be detailed Markdown files that AI turns into code.

So our understanding moves from the C/Python level to the Markdown specification level. We still need to understand exactly what we're building and why.

The hard part was never translating the spec into machine code. The hard part is defining the right spec. That is still the job of the software engineer.

> decades later they will suffer arteriosclerosis and cardiovascular problems.

That’s quite a claim, and the opposite of my experience. What’s your evidence?

I’ve seen long-term vegans develop health problems and improve after switching to a low-carb, meat-based diet. I’ve also seen the opposite.

I was vegetarian for over a year and ended up gaining weight and feeling like crap. I switched to keto, and now I am healthy and have lots of energy. It seems to work for me, but that doesn’t mean it’s right for you.

Maybe the answer is to figure out what works for you and ignore the vegan/keto propaganda and influencers.

> Make it make sense.

Easy. In the US, you get the justice you can afford. Spending money on lawyers, lobbyists, and “campaign contributions” to people in power can make a big difference.

And the AI companies have very deep pockets.

I’ve loved programming for more than 40 years, and now I enjoy it even more when I direct Claude to write the code.

That tells me something about what I’ve really loved about programming all along. It was never the act of typing code one character at a time. It’s understanding the problem and working with the higher-level ideas, algorithms, abstractions, and solutions needed to turn an idea into working software.

LLMs let me spend more of my time on exactly that.

It always amazes me how some vegans go into a frenzy over animals being slaughtered for meat while completely ignoring the bloodbath caused by farming practices used to produce vegan-friendly crops. Not to mention the pesticides that damage ecosystems and contaminate groundwater. The selective outrage and hypocrisy are jaw-dropping.

Have you ever seen a combine harvester kill animals sheltering in a crop during harvest, or considered how many animals are killed by the pesticides used in crop production?

The problem isn’t the law. The problem is that wealthy companies can use lobbying, political donations, and influence to avoid or weaken enforcement. In other words, the problem is corruption.

If you or your software hacks a website, the law will hit you hard. But powerful AI companies have the money and influence to avoid the same consequences.

The US administration just needs to enforce existing laws equally. Do that, and AI companies will suddenly have a very strong incentive to take safety seriously.

Yep, because private enterprises never go bankrupt, commit fraud, mistreat customers, bribe politicians, put profit ahead of safety, hack websites, etc. etc.

What world are you living in?

Not my experience at all (30+ years of pro experience). I love the fact that LLMs can take care of the low level coding while I focus on the fun stuff.

Hey, no problem, and thanks for the level-headed response. We need more of that in this world.

And yes, I agree that we still don’t know the long-term consequences of AI. But current LLMs are nowhere near good enough at judgment or new ideas to replace most jobs.

I expect they’ll mainly automate parts of jobs rather than replace them entirely. That’s exactly what I’m seeing as a software developer. I use an LLM to write my code, but it only works well when I apply good judgment and shape the context and prompts to remove ambiguity and then carefully review the results.

I hope the threat of LLMs making much of our existing software dramatically easier to hack will finally push the software community toward formally verifying security-critical software.

Ironically, LLMs are also dramatically lowering the cost of doing exactly that.

Formal verification isn't a panacea, of course. You can prove the wrong specification correct. But for the small, security-critical foundations that everything else depends on, I think the economics are becoming much more compelling.

My take: I don't think recursive self-improvement (RSI) will happen with current LLM architectures, even if they become 1000x "smarter".

RSI requires more than being better at solving problems. An AI needs to decide what is worth improving, come up with genuinely new ideas, and decide which ideas are worth pursuing.

Current LLMs are still weak at two of these: judgement (what matters) and new ideas. They're very good at combining and developing existing ideas, but that's different from independently finding important new directions.

That's also why I don't think most knowledge workers can simply be replaced by LLMs. Many tasks can be automated, but the most important ones can't.

I don't think this article adds anything that hasn't already been discussed to death on HN.

LLMs are powerful new tools, and we're all figuring out how to use them effectively. Some developers are getting great results, others aren't. Some instinctively love them, others hate them. Pretty much like every other developer tool HN has argued about, from Lisp to Emacs.

And yes, management often doesn't get it. But that's hardly new either.

Thanks for the detailed reply. It’s always interesting to hear about real experiences from a different software domain.

So it sounds like the main problem is the under-specified hardware. I can definitely see how that would limit the usefulness of LLMs.

I used to work on consoles, including the N64, which are fairly low-level compared with a typical Windows, Linux, or Mac application. But you’re obviously working much closer to the hardware, which is a completely different world altogether.

> nothing short of being a paper pusher

Hey, show a little respect for the paper pushers. Without them, civilization would collapse. And that’s not even a joke. :)

> I dare say even likely that you are missing nuance and observing bias from your bubble

I hope you’re not assuming you’re immune to being in a biased bubble yourself.

I’m just honestly sharing my experience. Obviously, your experience has been different, but both experiences can be true at the same time.

> YouTube recommends videos designed to keep us watching

Yes, but you don’t have to subscribe to what it suggests. I keep a carefully curated list of subscriptions and very rarely look at recommendations. That gives me a high-quality filter without all the algorithmic BS.

I disagree. Software developers seem to have adapted to AI much faster than the maths community.

There are certainly still strong opponents of AI, but they seem to be a small, vocal minority. Most developers seem to have simply adapted and moved on.

> Working on other people’s tasks with AI agents is nothing short of being a paper pusher.

Not my experience at all. I love not having to do the low-level code writing. I’ve been doing that for 30+ years, so I’m more than happy to let the AI handle that part now!

Demand for entry-level software engineers in the US is down, while demand for experienced engineers is up.

That makes sense given what LLMs are good at and what they’re not.

I work on proprietary software, so no, that’s not really a fair question to ask.

However, if you’re willing to share how you’re prompting the AI and some examples of where you think the results are poor, we might be able to help identify what’s causing the difference. I’d be genuinely interested in understanding why we’re getting such different results.

Rather than assuming one of us must be wrong, it would be more useful to compare approaches and see what we can learn from each other.

LLMs are great at pattern matching and “filling in the gaps” based on learned patterns. However, they are still really bad at (1) deciding what is valuable or worth doing and (2) coming up with truly new ideas.

So I don't see much benefit in this approach compared with human-in-the-loop development. It basically turns development into a waterfall process, with the human deciding what to build up front, rather than the more agile approach where the human and LLM continuously decide what to do next.

What an excellent read and project! This is why I read HN :)

My experience as well, working on very large-scale C++ code.

Thanks for adding such a thoughtful and level-headed comment to the discussion.