This has happened to me with Rust. I read the book, did the interactive tutorials and then started on a small project and couldn’t really code anything!
This is familiar territory for me as I prefer to do a breadth first survey of something I’m trying to learn before digging in to an actual project with depth to make all the mistakes and blunders that make you actually good at something.
AI can still be a huge help here. I don’t really use AI to write the actual code. I will be tooting along, in the flow state, and then bang off a question or request like “how would you write this block of code” “how does this API work?”, and then just take what it says under advisement.
I find as soon as I start accepting what the AI says without critical thinking or just hitting the “apply block” button, it’s time to take a break. I use CoPilot and I barely touch 40% token spend per month and to me it’s the best way to work, I am always engaged, have a full mental model of what I’m working on, and have someone pushing back and criticizing what I’m writing. This gets you out of tutorial hell, but keeps you motivated because you’re not always staring at a blank code window wondering where to start.
I’ve been using Kagi for about a year now and I genuinely get worried that there are no alternatives if it goes out of business. The results are extremely good, especially in the last 4 months. I like their opt in AI summary as well, just add a question mark at the end.
I pay for a lot of things that are free from google/big tech, I’m happy to watch the advertisement driven web implode on itself so we can go back to the idea of a consumer paying a company for a quality product, monetizing peoples attention has been a huge detriment to society.
The foundation points out something that has always been true, but AI has really brought to the forefront, that any contributor, including AI, could potentially not be relied on to maintain this patch in the future.
This is the core of the issue, not that someone uses AI, but that it’s just one of many smells a patch can have that indicates someone doesn’t understand what they’re submitting. You could be breaking variable naming conventions, changing APIs you shouldn’t, making amateur language mistakes, all indicate that yes, maybe the patch does work, but that there are other good reasons to reject it.
A way around this might be to mark a PR as rejected because of AI and then ask the author to point out some part of it they’re particularly proud of and explain in their own words, not a wall of AI text, what this does and why they like it. Just something where the author has to show that they have something an AI can’t, namely taste and an opinion.
Right. Reviewers still have the advantage of being able to spot AI text because it's often overtly different.
I just meant to say that, if you prompt ai "what would a human be proud of having written this code" you'll get an answer. They're not categorically incapable of fabricating an "opinion", they're just trained not to express one by default.
I’m teasing a bit, yes your point is well made, you could just prompt around it. I think the bigger picture is there isn’t going to be some magical panacea that fixes this, we’ll just have to accept some sort of Swiss cheese model, like in the world of aviation safety. This “test” could be just be another layer in the stack.
"(It’s famously not well capable of sounding human)"
Rather than a binary, I prefer to measure the question of "how much text does it take to be reasonably sure that it's an AI?"
By that metric, it is getting better at a reasonable pace. People are also getting better at prompting their AIs to write in something other than the default LLM style. If you think you're good at picking up that style, you probably are. But it's a lot harder to pick up AIs when they're fed a style sample. You wrote in the default AI style, and yeah, most of us have twigged to that by the end of your couple of short sentences. But feed the AI a style sample and it can definitely make it two sentences without every one realizing it's AI.
I think you first need to make an assumption of good faith on the submitter’s part. Yes, someone actively hostile to a repo and trying to sneak in changes will be successful, they can just prompt around the challenge.
I think the bigger picture is there won’t be one catch-all solution and we’ll need to embrace the Swiss cheese model from the world of aviation safety, this is just a suggestion for one layer.
It raises the stakes though. Getting challenged for an AI slop PR isn’t great, but ok, try and redeem yourself. Getting caught trying to cheat that challenge, you might as well just close down your account, like what is the point of even spending tokens to do this? These slop PRs are just people trying to pad their GitHub profiles.
I mean, yeah, you can give it style guidelines, but if someone's goal is "mislead people into thinking a person wrote this" then they should really be reevaluating their values.
One of the early selling points of LLMs was their ability to mimic styles. I haven't heard about that for a while though. Wouldn't that at least obscure the classic tells, if not eliminate them? I think the reason most (obvious) AI output is obvious is because people don't bother to hide it.
It's funny how identifiable AI text is. I usually know within like 10 seconds of watching a youtube video if the script is AI generated and it's a huge turnoff.
Plus the video generation, you basically know from the thumbnail. (Why do so many AI generated videos have a weird unnecessary film grain?!)
This could easily be circumvented by having the AI generate an explanation of the chosen portion of code and have the human rewrite it in their own words. It is much easier, imo, to swap clauses and put synonyms in place of other words within existing writing then it is to synthesise new text.
Literally installing Arch right now as I read this. Same deal, been using windows as my daily driver all my life but have been running Linux servers since late 90s (and did daily drive RedHat back in the day).
My work env was just VS Code + WSL and I realized the most pain points came from using Explorer and trying to admin the machine with the fractured landscape of sys admin tools. For me it became very obvious that windows is only going to get more bloated and less “my” machine going forward, why stay on this platform if I’m already spending most of my day in a Linux environment (that I’m already familiar with).
Last "job" was a startup with two other people and it was so great having others to bounce ideas off of and share the ups and downs. Building something together is a fundamental human joy IMO. We sold that and I've been working solo for the last 4 years.
It's lonely, and these have helped me.
- Making a point to physically get out of the house and ideally meet up with someone for a coffee or lunch or something.
- Being part of an active group chat.
- Podcasts
- Working out of a library or bustling cafe, just physically being around other people.
- Working out and maintaining a good wake/sleep cycle.
I also have some great hobbies (flying, scuba, rock climbing) with great communities that keep me socialized.
For anyone about to quit their job and try the solo dev thing, I always say mental well being has to be a "remembered priority". You're never going to have work/life balance like you did at an office job and the first few years are incredibly tough, so you need to plan for that. Marathon, not a race etc etc.
I think we'll start to see AI as any other tool that can atrophy your natural faculties. You can use a wheelchair to get everywhere, but your leg muscles will start to wither, but a wheeled vehicle for going longer distances is a genuinely useful tool.
Reaching for AI as a _substitute_ for thinking is bad, but reaching for it as a tool to assist thinking is good; you just need to be honest about whether it's your brain in the driver's seat or the chat bot.
> Reaching for AI as a _substitute_ for thinking is bad, but reaching for it as a tool to assist thinking is good; you just need to be honest about whether it's your brain in the driver's seat or the chat bot.
I think this is generally true, but human nature being what it is, the vast majority of people will use AI as a substitute for thinking rather than a tool to assist thinking. You can already see this from casual observation of today's AI users.
As I've grown older, I've noticed that more often than not, when someone says something to effect of, "Thing X can cause problems, but is great if used properly", you can be almost 100% certain that Thing X is going to cause very large problems and practically no one is going to use it correctly.
Unfortunately, "X is just a tool and is super useful when used properly, all things are both bad in excess and good in moderation, what you gonna do?" is exactly the type of conclusion that a chat bot is likely to reach. Doesn't really say anything, appears to express sophistication and wisdom by being more "nuanced" than an actual position, demands nothing of your audience, not likely to get downvotes on social media, etc.
I don't think the "tool weakening" discourse is strong enough: it overlooks the aversarial nature of the modern internet. There are humans actively intending to weaken you for various reasons, either to sell you stuff or to weaken you ideologically by making you hate other humans such as in this comment https://news.ycombinator.com/item?id=45848215
Yes, IMO this is going to start becoming visible sooner rather than later. College students that defer to ChatGPT to form arguments for them are going to graduate, sit for an in-person job interview and discover they haven't had to think fast, with their own brain, in years. It won't be pretty.
But who is going to crack first? Will job applicants somehow remove their borg implants and learn to think on their own? Or will businesses give in and admit that nobody can think for themselves anymore, allowing applicants to use ChatGPT during their interviews (knowing that they're probably going to need to use it on the job, too).
Steve Jobs "bicycle for the mind" analogy is more potent than I initially thought.
When got past the bicycle phase where we augment our body with technology but still leave room for our body to improve. We got into the automobile phase where only the goal matter and the body is not participating (and improving) anymore.
(well, except maybe for F1 which are bona fide athlete, but your average driver in a traffic jam is most certainly not a F1 driver)
> You can use a wheelchair to get everywhere, but your leg muscles will start to wither
Although it's not to the same degree of atrophy, I've been thinking of cars the same way. They're too easy to use once you have them, just press the pedal, so you stop walking and everything becomes a matter of driving distance, which makes it acceptable to distribute commercial activity in stupid little pockets of car destinations and avenues separated from each other by noise, pollution, and danger. People may not physically atrophy to the point of having no leg muscles, but their tolerance for walking a km becomes much more strained and their appreciation for investment in public transport lessens. They don't see or speak to people as much because they're always in their portable silo. You burn less calories, it's easier to gain weight, and people discount the value in having a gym within walking distance.
It's tough to reconcile it with being a functional tool, because although I could conceivably use it as one and buy one, I know that it can become an addiction.
Sure thing, sometimes it's fun! I suppose that's because I'm a thrillseeker. Let's see how fast it gets up to is a test. That wouldn't be the case if I were driving.
>You can use a wheelchair to get everywhere, but your leg muscles will start to wither
I don't know if they still exists but there was a (Dutch, I think) company that makes non-electric lifts and tools for elderly people at home that require muscle effort from the people using them. Purely augmentative tools that don't work without input.
And that is how it needs to be. Framing this as choice is already wrong. Any tool that is agnostic or conducive to forfeiting agency will be used in wrong ways. It's not enough to make the healthy part optional, your brain will not be in the driver seat given how human beings work.
People in Japan are slim with no spending while people in the US remain obese while spending billions exactly because to the Japanese this isn't a choice, it's one the environment makes for them. If you rely on people "keeping themselves honest" you've already lost.
"For this invention will produce forgetfulness in the minds of those who learn to use it, because they will not practice their memory. Their trust in writing, produced by external characters which are no part of themselves, will discourage the use of their own memory within them."
> This discourse is as old as humanity. Every tool makes us stronger but also paradoxically weaker.
Of course that statement is true for every tool, but what's missing from the discussion is whether the trade off is worth it. Even truly terrible things have benefits. Smoking cigarettes makes it easier to maintain a healthy weight, this is well documented. Smoking has also been shown to reduce anxiety in some people. The negative consequences that cigarettes introduce, however, are so horrific that no one in their right mind would recommend that someone take up smoking, even if there are some demonstrable benefits to it.
I do not believe generative AI has any benefits whatsoever! If it does, I have yet to discover them. It may be a benefit to everyone here which is great, but i would say that it's better to spare it.
Lol I refuse to use it most of the time. If I want to test it, I do. I probably shouldn't do this, because it's keeping all data! It's good for some casual fun sometimes, I usually navigate away after I'm finished though.
Curious if we could test/compare (popluation-level) memory skills before/after writing was introduced to the population.
I want to say "I remember things better when I write them down", and because I think I'm a smart person I think my memory is good.
I don't know how well I'd remember things if I'd spent a large portion of my life building memorization skills. Maybe I could be 100x better at memory if I exercised it more?
You could theoretically do it with uncontacted tribes that don't practice writing, but in practice it'd be very very hard to establish a scenario to administer such a test, especially ethically.
Even then it would be a bad study because you'd have an extremely narrow biased sample with a very specific culture. It'd be impossible to separate out cultural or genetic differences.
there seems to be a parallel with the industrial revolution - being fit and having muscles used to be the norm when everyone worked the fields all day. but now that grocery stores and sedentary jobs have made exercise optional. so choosing to pursue fitness signals to others that one is disciplined, takes pride in their appearance, etc.
i can see the next couple generations of AI agents causing the same effect on reading, critical thinking, and intelligence in general. thinking is no longer necessary with AI agents, so maybe cultivating one's ability to think will become optional/personal pursuits which send similar signals.
I noticed this in myself and had a pang of disgust at myself. I used to write almost daily, but with the baby we've had there hasn't been time. So recently when I was thinking of getting back on it, I went testing several writing helpers that are LLM-powered. I think it took me a few days to realise I was only doing it because it's easier and “everyone is doing it”. Like, I write for pleasure, why the hell do I need to automate part of the process?
Design is one of those things that succeeds or fails in subtlety and both are difficult to quantify and back propagate through any sort of process, let alone training a model. The same way we figured out that the microwave can make approximations to good food quickly, so too shall we see that AI can do the same with tasks that rely heavily on a connection to people's aesthetics.
This is familiar territory for me as I prefer to do a breadth first survey of something I’m trying to learn before digging in to an actual project with depth to make all the mistakes and blunders that make you actually good at something.
AI can still be a huge help here. I don’t really use AI to write the actual code. I will be tooting along, in the flow state, and then bang off a question or request like “how would you write this block of code” “how does this API work?”, and then just take what it says under advisement.
I find as soon as I start accepting what the AI says without critical thinking or just hitting the “apply block” button, it’s time to take a break. I use CoPilot and I barely touch 40% token spend per month and to me it’s the best way to work, I am always engaged, have a full mental model of what I’m working on, and have someone pushing back and criticizing what I’m writing. This gets you out of tutorial hell, but keeps you motivated because you’re not always staring at a blank code window wondering where to start.
reply