deterministic

I am seeing the complete opposite. 200k senior engineers can now do large scale projects way beyond what they could do just a year ago. My company is now able to implement stuff that we used to only dream about. Having 200k senior engineers who master AI coding is a true super power.

> They built a game engine to render text for cripes sake!

Nope they didn't. I recommend taking a look at a real game engine (Unreal or Unity) and compare.

I personally wrote two game engines used for commercial games. Which included rendering text. It is not hard.

I trust the C++ committee to introduce new features in the most convoluted way possible, then spend the next 20 years trying to fix it, while adding even more syntax that makes my eyes hurt.

Case in point: templates. They are essentially a pure functional programming language embedded inside C++, expressed in a verbose syntax that barely resembles the rest of the language, and somehow makes even Java look concise.

It has been a slow-motion train wreck, with one questionable design decision after another. And a perfect example of why design by committee often leads to unnecessary complexity.

That hasn't been my experience at all. We've been using a custom code generator for years to build a large number of business-critical applications.

The generator takes a single specification and produces everything needed for the server, client, and databases (SQLite, Oracle, in-memory, etc.) to stay perfectly in sync.

It has worked really well for us and has been a huge productivity boost.

Yep learning is really a participation sport not a spectator sport. In other words, you only really learn by creating something not just reading about how to create something.

Agree. In my experience, the only way to overcome anxiety is to push yourself to do the thing anyway.

The more you safely do something that scares you, the more your brain learns that it's safe.

The key is to break it down into small steps. Each step feels less intimidating and the brain learns bit by bit not to feel anxiety anymore.

It most definitely worked for me.

Not verifying the generated code is a really bad idea.

I feel the complete opposite. I am much more relaxed now since I can be way more effective while actually working less.

The less people know the simpler they think things are.

You should make $ bets with them on whether you will still have a job 2 years from now. An easy way to make money and teach them a lesson :)

Thank you for posting such balanced and level headed comments. We need more of that on HN.

I completely agree. The more experienced you are, the more you get out of AI.

A friend of mine is super smart, but he's not a software developer. He tried to use AI to build software. And it worked well for small projects, but he always reached a point where the AI could no longer fix bugs or add new features.

The problems were often simple. An experienced developer could fix them with a few minutes of manual editing. Sometimes it was just an incorrect > in an XML file, but the AI would get completely stuck.

So IMHO AI works best when you know how to guide it and when to step in yourself.

I've read a lot of comments about using AI for coding, and my experience has been very different.

I work on large C++ applications used by international airlines. If this software failed, it would make national headlines.

Claude Code with Opus 4.8 is great at handling the boring code I don't want to write myself. It gets it right almost every time.

However I still review every change and test everything before committing.

Trust, but verify.

Master's here. I already knew how to program before going to University. However I learned a lot about how to think more abstractly/mathematically about software. It has helped me a ton in my industry work and given me a clear advantage.

A typical problem with software is that development continues after the software has already reached its peak. Instead of making it better, new features and changes gradually make it worse. Eventually, enough users leave that it is no longer worth paying a team to keep working on it. And then the cycle starts again.

Cherry-picking individual technologies (such as jet engines) doesn't really say much. You could argue that companies like ASML and Rolls-Royce (jet engines) are evidence that Europe knows how to innovate and the US doesn't. That Airbus overtaking Boeing in a market once completely dominated by Boeing shows the US has lost its edge. That the European-designed ARM architecture winning the mobile phone wars shows the US has lost its chip design advantage. And so on.

But there are obvious counterarguments if you cherry-pick technologies where the US currently leads — Google Search, AI, and so on.

So I would be really careful extracting any kind of simple "truth" from examples like these. Different countries have different advantages, and those advantages shift over time. That's it.

Check out seL4 and CompCert for real-world examples of complex software proven correct and used in industry.

Maybe you could have generated random test input, generated the trees, and then converted the trees back to the input, checking if they match?

30+ years of experience here writing everything from JIT high-performance compilers to professional game engines to mission critical biz servers:

Claude Code is excellent at writing (C++) code. Especially multi-threading code. However I still need to verify everything because it will occasionally make mistakes.

I am really curious about this because my experience is completely different. Can you elaborate a bit more on how you use LLM's for coding? Are you doing one-shot coding? Vibe coding (not looking at code)? Small focused one-step-at-a-time with yourself verifying? Anything else?

I agree. Claude Code (for example) writes excellent code as long as the prompt is focused and clear.

Not my experience at all (30+ years of experience as a professional software developer).

I currently work on a very large (multi-million lines) C++ code base and use Claude Code for the following:

1. I run Claude Code in the background finding bugs, while busy implementing new features.

It has so far found and fixed 100+ minor issues and a few major could-crash-the-server-in-production issues that I was really happy to get fixed before a customer found it.

2. I use Claude Code to do massive refactorings that would be really tedious to do by hand.

I love the fact that it updates not just the code but also comments and documentation.

3. I use Claude Code to implement features that I frankly am too bored to implement by hand.

Excellent! Less boredom at work!

4. I use Claude Code to generate tools to analyse the code and/or generate code.

This is a massive quality improvement. It allows me to make the code more maintainable and simpler while adding more features.

However I always always always check and verify every change before committing to git. Because the LLM it will occasionally add features or bugs I don't want.

The end result is really positive. Lots of productivity improvements, less tedious can't-be-bothered work, lots of bug fixes that otherwise would have triggered production issues etc. etc. while 100% maintaining my understanding of the code and all changes.

Love it.

The public school system in most countries has been spectacularly successful. In fact, much of the West's success is built on it.

Having said that, people are different. They have different IQs, interests, abilities, motivation, and levels of family support. No education system will work equally well for everyone.

That isn't a problem with today's public school system. It's a problem that any education system would face. Someone with an IQ of 70 will never understand concepts like compound interest, no matter how much time, money, or tutoring is provided. While a smart and motivated person will usually learn despite a mediocre school system.

This is why every attempt to "fix" the public school system has failed spectacularly.

It's amazing how badly written most manuals are.

Communism is authoritarian left. None of the AI's match that.

Non-paywall version?

It turned out in practice that a Lisp Machine was basically a Lisp interpreter implemented in hardware. A better and much more pragmatic solution was to compile Lisp to efficient machine code and then run it on any traditional CPU. Which is what we do today.

That is a very unusual experience and not at all what happens in general. Quite the opposite. The more skilled you are, the more effective you can apply AI to solve problems. I suspect your company is destined to hit a maintenance/AI slop wall at some point in the future. My recommendation is to polish your CV and find another job ASAP.

I can't remember where I read the statistics. But apparently helping with safe drugs/housing removed 80% of homeless from the streets. However there were still 20% that couldn't be helped no matter what they did. That is of course not an argument against helping. 80% is amazing!

> money doesn’t buy happiness

Not true. Money does buy happiness up to a point. Going from sleeping in the streets to having a place to stay, food, security, and medical services makes a big happiness difference.

However once the basics are there, additional money makes less of a difference (but not 0).

A friend of mine is an artist working in the games industry. So far nobody has been fired. They all simply started using AI to be more productive.

They care more about doing their job well than some artistic ideal that only works in practice if you don't have to care about things like food and shelter.