How ironic would it be if the AI productivity miracle ends up being negative growth because of the cost of cleaning up of slop?

Honestly I think it already is.

A lot of people who talk about massive gains seem to forget about code review.

Then in a company someone who has to review is is f*cked because that code is much more complex and takes much longer to review.

I tend to share the sentiment of the author.

I think that coding assistants tend to quite good as long as what you ask is close to the training data.

Anything novel and the quality if falling off rapidly.

So, if you are like Antirez and ask for a Linenoize improvement that has already be seen many times by the LLM at training time, the result will seem magical, but that is largely an illusion, IMO.

Heads up that this is "more true" for non-reasoning LLMs. Reasoning gives an LLM a lot more runway to respond to out of distribution inputs by devoting more compute on understanding the code it's changing, and to play with ideas on how to change it before it commits.

But how many times a week are you truely doing something novel? A thing that nobody in the world (or the LLM training data) has done before?

Argh.

I may suffer from some kind of PTSD here, but after reading a few lines I can't help but see the patterns of LLM style of writing everywhere in this article.

Holy cows, same. But it's not PTSD. People think you can tell because of ternary rhythm, —, big-words and other shenanigans. These are tactics over strategy, obsessing over the minutiae. But I think what you're feeling (and what I sure fucking am) is a lack of voice.

This writing says something, has a point, and you could even say it has a correct way to get to the point, but it lacks any voice, any personality. I may be wrong -I stopped reading midway- but I really don't think so.

same. I just seem to automatically tune out

I've been following the development of Ghostty for a while and while I have the feeling that there is a bit of over-engineering in this project, I find this kind of bug post mortem to be extremely valuable for anyone in love with the craft.

Over-engineered in what way?

It’s just a feeling, man.

Having to introduce a new language stack to distributioms just to be able to build a terminal emulator is what I would consider over-engineering.

So anything that uses a less popular language is considered over engineering? Distros support lots of different languages already and there are likely other packages built with zig already.

A 50-ish MB build time dependency that doesn't need any special privileges or installation to run? That's over engineering? A binary release of just CMake is bigger than all of Zig.

The technology of air travel may seem counter intuitive when your frame of reference is the Moore's Law.

But in practice, what happened with semiconductors is the exception, not the rule.

We are still often making wild predictions about the future of technology based on some kind of exponential take-off, it may turn out to be a lot more constrained by physics and energy density.

Supersonic commercial air transport is one such technology, possible and proven, yet not viable.

Mars colonies or interstellar travel could be in a similar bucket.

Except the title, a large part of this article seems to have been written by an LLM.

Crazy that you are the only comment that noticed this. I stopped reading at,

> Translation: Building airplane engines is hard

There are many hints previous to that, but that gave it away for me. If I want LLM output I’ll request it from the model myself, thanks.

I am pretty sure that many car brands will discover that this can be a selling point and reconsider the cost/ux tradeoff.

How much of this was made with LLM?

Why does it matter?

Because when a project is done in 10 minutes by llm - it will be abandoned in a week.

When a person intentionally does it and spends a month or two - they far more likely will support it as they created this project with some intention in the first place.

With llms this is not the case

Why are you entitled to ongoing support of a free tool?

How long are you entitled to such support?

What does “support” mean to you, exactly?

If the tool works for you already, why do you need support for it?

A bug from slop could cost $10K

So could a bug introduced by a human being. What's the difference?

Accountability is the difference.

An LLM is just an agent. The principal is held accountable. There’s nothing really all that novel here from a liability perspective.

That was my point exactly. I just didn’t write it as precisely as you.

Then I don’t understand. My point was that it doesn’t matter whether the machine or the human actually wrote the code; liability for any injury ultimately remains with the human that put the agent to work. Similarly, if a developer at a company wrote code that injured you, and she wrote that code at the direction of the company, you don’t sue the developer, you sue the company.

How exactly do end users hold AWS devs / AWS LLMs accountable

The human

How much would a bug from a human cost?

I’d be willing to bet the classes of bugs introduced would be different for humans vs LLMs. You’d probably see fewer low level bugs (such as off-by-one bugs), but more cases where the business logic is incorrect or other higher concerns are incorrect.

Text can be surprisingly immersive and rich, often surpassing the most complex VR experiences.

It is amazing what we can do with a few strings of symbols, thanks to the fact that we all learn to decode them almost for free.

The oldest and most important technology indeed.

I agree, a trailer should focus on emotional reaction, not a simple display of features or quirkiness (1400 puzzles,10 years of dev). Besides, the voices and writing are generic and maybe even AI generated. The witness had a really good promotion canpaign beautiful and intriguing.

> not a simple display of features or quirkiness (1400 puzzles, 10 years of dev)

I think a 'number of features' metric can work but only for players that already know and like your game, where an expansion with 'Five exciting new areas' is understood as something that they'd enjoy, and I agree it feels odd for a new IP.

Similarly, saying how many years it took isn't remotely a selling point for a new player. If you'd been following the development process then you probably wouldn't care, and if you hadn't you also probably wouldn't care.

It does seem awkward to have to design a trailer for a pure puzzle game, something that essentially relies on things going on inside a player's mind for fun, which by definition won't be visible.

Baba Is You did have something you can show potential players, but I'm not sure there's a trailer that could convey The Witness' 'Oh, I wonder if I can...' moment as it's a very internal experience that comes from playing enough to get to that point.

The Witness was, however, visually beautiful (IMO) and its symbol-based language let the trailer keep an element of mystery and intrigue. Order of the Sinking Star, while potentially also a fantastic puzzle game, seems to not be able to hide anything by nature of it being very clearly a Sokoban-like. Even if there are as-yet-unseen depths to how it treats the Sokoban format, the trailer needs something to work with, and while I think it also looks lovely it perhaps doesn't have the The Witness visual appeal or mystery to draw people in.

I guess the idea is that Blow has a built-in audience and is appealing to them. "1400 puzzles from the maker of Braid/Witness? Wow!"

Blow falls into a classic engineering mistake of marketing the challenge or effort to make something (audio logs everywhere) and not the end experience.

Note that when masters like Steve Jobs do it, they mention it very quickly, or they mention the ideals of craftsmen ship, rather than the actual process.

Essentially Braid Anniversary edition. Huge effort for someone to just say “oh so just Braid Remastered?” In general from his streams you learn that so much goes wrong during the slog that is video game development. Hire failures. Contractors billing $$$ and writing 1-2 LOC. Devs rage quitting. Platform optimization. Even a suicide (not just an employee but someone close). They do all this work and dropped the ball on marketing despite betting the future of the company on its reception, but even failing that I think it was clear to a normal person (which can be hard to reach out to and interact with) that there was no appetite for this game.