I've been a game dev for 25 years, and I'm not so full of hubris to say that no LLM will ever be as good at coding as me.

But most game code isn't public domain, so there's not going to be a lot of training on it. There's a lot of weird stuff that we do at the behest of designers that nobody would ever think to do (whether that's because it's a terrible idea or a brilliant idea is really only knowable after people play the game).

The devil has always been in the details, and so much of good game development is about good human communication and understanding what makes a fun game. Sometimes you just don't know. You're 3 years into a game and there's all this tech and it still isn't fun, and you're trying to figure out the special sauce that will engage people and it's hard. And the way we usually get through it is we play a lot of games and all of us come in with our own idea of fun and we take chances based on things that we personally enjoy. Every professional designer I know plays games as homework. (Some companies help defray the cost, some just demand that they do it as a condition of their employment; it doesn't much matter, designers are gonna play games no matter what anyone tells them.)

And so I'm relatively confident that games will mostly be a collaborative effort because programmers ALSO play games and ALSO bring that experience to the table. As much as designers and programmers will complain about one another, we really do feed off of the contention. Programmers provide infrastructure, but also BOUNDARIES. We know what's slow and what's fast and what shortcuts we can take. And hey, maybe that will all be possible with an AI in the future too, but I have my doubts.

But I had Gemini fix a bunch of garbage problems with some elisp that have been plaguing me for a few years that I never would've bothered to look into because my setup was good enough. I listened to one programmer talk about how he tackled a problem of (human language) translation that he never would have attempted because the task was too large and too tedious.

I actually don't think that LLMs will be making programmers meaningfully more productive for a long while. I don't write enough so-called boilerplate code for it to save me any time there. But I'll definitely take on some tasks that never would have gotten done at all. I think we might see more small websites pop up, for instance. I've never bothered to learn anything about web dev after the HTML/Geocities/webring days, but maybe I'll do something now that LLMs can walk me through it. Small automation tasks, writing scripts. Hobbyist stuff for people that don't want another hobby.

There's already a lot of human-generated slop code out there, and maybe the folks that write that stuff are going to be replaced. I think there are really technical or esoteric or experiential programming jobs out there that won't be, though. I just have more time to do my actual work because all of the dumb little distracting tasks that I used to spend a half day on are now 10 minute LLM chats.

Actually, a second comment to expound a little:

Few people commit crimes knowing that they'll get caught. I would daresay that nearly 100% of criminals believe that they'll get away with it.

If you think you're going to get away with a crime, you're not going to consider the potential consequences. Alternately, some people commit certain kinds of crimes because they believe they've got a moral cause to do so (some kinds of protests, refusing the draft, etc.)

Given that, the severity of the penalty probably doesn't enter into their minds much. Those 11 people thought they could get away with their crimes long enough that the death penalty (which they surely knew was on the table) wouldn't matter.

Indeed, if you actually go and look up the research on it, it doesn't appear to have a meaningful deterrent effect over and above life imprisonment.

So again, to reiterate: you don't know what a deterrent is.

You don't know what is.

The death penalty isn't a meaningful deterrent--we still have a lot of crime. We still have these 11 people perpetrating crime until they're caught and killed by the state. This assumes that they're actually the guilty parties and not some fall-guys for the real criminals.

Punishment is the *worst* way to *prevent* crime because it relies on:
a) someone actually committing the crime (that is, the harm is done)
b) someone caring enough to solve the crime
c) the system prosecuting the crime

The reason the whole carceral system sucks is because harm is done to victims and the only "justice" they get is the perpetrator maybe not doing more crimes. They don't get compensation or therapy or anything. The state fails them utterly. I'm sure these 11 people being dead (assuming, again, that they were the actually guilty parties) gives the victims some measure of comfort, but it sounds like their lives were still pretty much ruined, so I dunno.

Compare to Scandinavian countries where they're closing prisons down because they don't have prisoners to put in them because they simply have less crime, and when someone DOES commit a crime, they rehabilitate them.

I don't trust the criminal justice system to do the right thing in MOST cases. It's shown itself to be subject to corruption and the worst aspects of crony capitalism the same as every other industry. I suppose these days we need to be worried about governments executing people in the streets with no trial, but I'm not convinced that the show of a trial is much help either.

Games, like movies and music and most other art forms, have been political for decades. The problem is mostly the second thing: the mercenaries and bad CEOs.

There's plenty of 'politics' even in fun movies, but you don't notice, because they're good. Make a bad game and everything sticks out like a sore thumb, and then it often gets blamed on the 'politics'. Most of the games that were 'ruined' by the political message would have been bad no matter what, because the devs weren't given the creative space to do better.

There are games that are practically ONLY politics ('Papers Please') comes to mind, but the games are so compelling, that's not necessarily what you notice. Games aren't purely about escapism any more than visual arts or music are, though we often go to them for that purpose. But as with most things, if you let a big corporation get their hands on it, they crank out the most profitable pablum they can and they don't really care what it says as long as their margins are big enough.

Every mid-level or higher studio will have at least one optimization person. You don't actually need a lot of them. Unreal's profiling tools are pretty decent, and a generic programmer can usually tackle the low-hanging fruit there. A lot of slow code is simple but unexpected stuff. For instance, Unreal's penalty for spawning an Actor is quite high, and you're generally discouraged from spawning a lot of them. If you fire a gun that uses projectile bullets, you may find the performance is bad. A simple pooling system will solve that problem. It's not an immediately obvious thing to do if you haven't seen it before, but it's also not a difficult problem to solve.

Unreal also has the problem of Blueprints--the visual programming language for designers/non-programmers/prototyping. It's very powerful, but very slow. If things get bad enough, you may have to take some of the blueprints and make them into native C++. But again, that doesn't take any super-special skill.

For the complicated engine stuff, depending on the size of the team, you might have one engine guy and one graphics guy, and that's about it. Bigger projects have more of these people, but fewer than you'd expect. Gameplay code takes by far most of the time and energy.

I dunno, what IS the chance? Do you have a good analysis/source, or is this just something you're abstractly worried about? What are the chances that the US government has a kill switch in US-made cars? They seem like a bigger threat to Canada than China right now, so should we stop buying American cars?

I'm more worried about the stuff that EV makers are TELLING US, like they might be recording us in our cars and uploading that data for customer tracking. I think it was Hyundai/Kia that basically said, "hey, if you have sex in your car and we hear it, that's on you."

I don't think Carney sold us out at all. This is the bare minimum.

If we're going to exist in a nominally capitalist society, we need to be able to reap the sole benefit of capitalism, which is competition making better products for less. Not only has the government propped up the auto industry for years with subsidies, tax breaks, and union weakening legislation, we get to pay more for our (huge, wasteful) vehicles to boot. It was all the other governments that sold us out. The amount of money that has gone into the auto industry for a piddly return is just enormous.

Let the Chinese in. Let them build cars here. I do not give a single shit about Ford or GM and their stock prices. If they can't hack it, gtfo. I love a union job, but they sure don't, and they've been destroying the unions here for years. We need cheaper EVs here as part of our energy transition, and all they want to do is build lifted full-sized pickup trucks that stick out of parking spaces. If they had properly committed to EVs here in Canada--I don't give a fuck what's happening in the USA--maybe they wouldn't be in this mess.

We're mired in a mess made by the corporatists that have run this country for the last 3 decades, and now they're in a bind. Maybe if the Chinese come in with cheaper cars they'll get their acts together. I absolutely refuse the propaganda that the Chinese are unfairly subsidizing their industry, they just subsidize their industries more effectively than we do. (For instance, they promise cheap electricity to companies if they're building certain things, so everyone jumps in to try and build the best product. The poor competitors are weeded out, the ones that are most efficient survive. This is in contrast to Canada just offering money and tax breaks and those companies ending up doing stock buybacks and executive compensation packages.)

Either the established car companies will survive or they won't, but that's not supposed to be my concern. If Canada wants to NATIONALIZE a car company to make sure there are always jobs, I'm okay with that too. Despite the obnoxious whining of industry executives, governments are often extremely good at running corporations because they're not concerned with profits for the sake of profits.

tl;dr Fuck NA car companies if they can't make it work under the current regime. They deserve nothing.

Fair enough! (Nice to see another Canadian here, too.) :)

This happened to me the other day too. I don't know why I get suckered into it.

I had a problem where my LSP was returning ??? at the beginning of completions; there were some non-UTF-8 characters in there and I didn't know where they came from. ChatGPT said it was a known problem from some Visual Studio compile flags when they get passed to Clang. In retrospect, that's stupid.

But it said it could solve the problem for me by intercepting the .rsp files, stripping the bad flags, and then passing them on. It wrote a little python script to launch the lsp and strip the flags.

I ran it and it seemed to work, but my testing protocol was wrong, and I was just seeing some correlation--the problem was intermittent, and me putting the script in place happened to coincide with some good responses from the LSP.

When I tried shutting down emacs and starting it again, it couldn't stay connected to the LSP through the script, and of course it couldn't. So the script was never really in play. After searching around for a while, it was a known error with a package called eglot-booster, and it was easy to fix by turning off a particular flag.

I have since instituted a rule where it MUST, WITHOUT FAIL, provide a citation for anything it tells me. If it can't find a citation, it MUST mark the section as unverifiable. That has helped a lot.

From my experimenting, Gemini is excellent at giving citations, but it gives up very easily if you tell it to do something and it can't do a complete job. I tried to have it summarize all the commands in a package that started with : (colon). It did an incomplete job, and when I asked it to do it again, it just told me where the documentation was. Claude did an excellent job. ChatGPT also did a partial job, but maintains conversational context better, so you can often get it to fix its own problems.

It may be just as good. It's Gemini, after all.

And I'm saying that if they can't get it the usual way, that MAY be a problem with the health system, not your friends. I was denied statins for YEARS because I was 'too young' to need them, even though my cholesterol was very high considering I was actively racing bikes.

It was only once I went to a cardiologist for something unrelated that he said, "absolutely not," and immediately put me on the statins, and I'm currently at the highest dose available. Despite the fact that I myself am nowhere near overweight *I* may end up on a GLP-1 drug because they also have better health outcomes for people like me. But the time between my first adverse cholesterol readings and the prescription of the statin was TEN YEARS.

Incidentally, I've never talked to a professional, non-sports dietician that wasn't completely useless. They asked me irrelevant questions about my diet and when confronted with the fact that they could do nothing, made up garbage that later, more expert sports dieticians told me not to do. Most of what a dietician does is exactly what we all know: smaller portions, more vegetables. Useless.

You are specifically talking about people who are somehow in need of the drug. They're overweight and sedentary by your OWN description. The reality is that in most cases, losing that weight, however they do it, will lead to better long-term outcomes. It doesn't matter that they're doing it in a way that you disapprove of, they'll probably live a healthier, longer life for it. If what you're concerned about is quality of life, drop this weird strict adherence to the right way to lose weight and get healthy. It is not more moral to suffer for your weight loss, that's what I tried to get across in my first post.

GLP-1 drugs have existed for decades and are extremely well tested, well tolerated and have more benefits than side effects. Even just from the perspective of side effects, it's almost certainly better to bear the ones associated with Ozempic rather than the ones associated with obesity, coronary artery disease or type 2 diabetes. From a public health perspective, to me, it's a no brainer. Loosen the restrictions, get people losing weight FIRST, and then try to fix the other problems with their lack of exercise and poor food choices. The other way around is a waste of time.

I hope you come back to check this.

So I'm close to 50 now, and I don't really care about my bf% anymore. I'm much higher than I used to be, but I really can't stress enough how healthy I am, fitness-wise. I'm a competitive masters swimmer. I cycled more than 5000km last year (135 sessions, about 250 hours worth) and while I only swam for 40h and 75km in 2025, I usually swim more like 250km a year, for 100+ hours. I lift weights to maintain my bone density, because cycling and swimming are both weight-bearing exercises.

I think your goals are GREAT, and I hope you accomplish them. I'm just at a time in my life where my goals are to extend my active lifetime and the body fat mostly doesn't slow me down. I'm almost certainly into the 20% range now, but people with this sort of bf% have LESS all-cause mortality. (Fat is protective in later life; getting sick will starve you and survival can often hinge on how your fat reserves hold up.)

Anyway, good luck with all of that! Those goals sound more than attainable, and it's good that you have a plan!

It's not clear what you mean by 'overuse', though.

Like I said, I have a friend who has never been overweight, let alone obese. She takes it 'off label' to quiet the food noise. She'll never lose any weight on it, but she doesn't need to. It's a mental health drug for her.

You're also describing people who claim they can't lose weight, and you blame their diet. But the implication here is that they SHOULD lose weight, they're just not doing it 'the right way'. But if their health outcomes will be better, it's not clear why you think this is 'overuse'.

Further, even if they DO eat a bad diet on Ozempic AND they lose weight, it necessarily means they're eating LESS than they would've without the Ozempic, so again, how does this count as overuse? Indeed, if their diet is bad and they're losing weight, it means that they've made their diet LESS BAD.

This is senseless moralizing about what the 'right' and 'wrong' way to lose weight/eat less is. The reality is that even if you're 100% correct about the people YOU know and they SHOULD move more and eat better, you don't know what the situation is for other people. While the folks you know may get on it and eat poorly and still lose weight and do nothing with that, SOME people will use that opportunity to get to the gym, to make their diets naturally healthier, etc.

You haven't made a compelling point as to why this is 'overuse'. The fact that people are getting it in shady ways is more of a condemnation of our health system than anything.

'More productive' HOW? More lines of code? More features completed? More bugs filed? More bugs fixed?

Unless I know exactly how he's measuring productivity, this means nothing at all. Maybe it's just 'more programmers are using AI', in which case he hit his metric, but I don't know how it translates into actual *work done*.

CEO slop, just like Bezos numbers.

This is overly reductive.

I was an amateur competitive cyclist and swimmer for years and years. It left me with an eating disorder and body dysmorphia for a long time, because I DO happen to be good at dieting and eating less. I don't actually recommend it to people, honestly.

But many people have something called 'food noise'. When they're not eating, they're always thinking about food. GLP-1 cuts the food noise down considerably. I have a friend who's never been fat, but she's microdosing GLP-1 because it helps get rid of the food noise that rattles around in her head all day long. If you're someone who has persistent food noise, it's HARD to not eat. It's miserable. Your physiology is telling you to do something and you're refusing.

I happen to be great at being hungry. I walked around with a persistent headache and a sick feeling all day long, but I'm pretty good at suffering, so I just did that. I'm 190cm tall and I was as low as 72kg at some point. It didn't make me THAT much faster as a cyclist. I would've been better off being 80kg and spending more time in the gym. Now I'm 95kg (220lbs) and I deadlift 400lbs (180kg).

So I've DONE THE TIME, and I will tell you that keeping weight off is hard even when you're disciplined. I weighed my food. I went to bed hungry. I looked at myself in the mirror and thought I was fat, because that's what doing these stupid things will do to your head. It took me years to get over my disordered eating and I don't know that I'll ever outrun the dysmorphia. I still ride and swim and lift, and I'm constantly annoyed at how little I like my body.

GLP-1 is great for people, and they should use it. I'm not sure why we spend time moralizing about this stuff. If you're obese or diabetic, GLP-1 will make you healthier and help you live longer. It'll help you lose enough weight that you might be able to start doing exercise that you want or spend time in the gym, but I don't think that's a mandatory requirement. I'm on statins because I'm genetically predisposed to very high cholesterol (and this was true even in my 20s and 30s, when I was at my lightest). Take the medication. That's what it's for. Are you gaining weight now that you've come off of GLP-1? Go back on it. We fuckin' did it, man, we made a pill that will make your life longer and healthier, and we're sitting here debating whether people should use them if they want/need to? What a waste of time.