"As of right now, however, not many people are using AI to code, despite the fact that it requires zero prior coding knowledge to produce fully functional projects. With AI, if you want to start a business, you can not only create your website with AI, but you can now run all your business processes — from expense tracking to payroll — with a set of prepackaged commands."
I'm a non-software engineer who has been using coding agents for months now both at work and for personal projects and gets a lot of value out of them, and I don't agree with this. Yes, you can create little programs and scripts reliably with AI, but I would not trust it to write enterprise-level code, and I especially wouldn't just ask Claude to run crucial parts of a small business as your link suggests. I would not recommend small business owners rely on AI to that degree.
Why not? For many reasons:
1. If something goes wrong, you as the business owner are liable rather than Anthropic.
2. Enterprise code is reliable, efficient, secure, and built for myriad edge cases in a way that a vibe coded app is very unlikely to be (and you wouldn't be qualified to evaluate whether it is or not).
3. Hallucinations have become a lot less common, but they do still happen, and especially for process like payroll where you need to be right 100% of the time, they're still not reliable enough.
I've been working with Claude inside VSCode a bit, over the last few months, and while it's extremely useful, sometimes it makes boneheaded errors. Worse, sometimes it makes more subtle mistakes that require careful review to catch and correct.
The possibility that Mythos will be able to set up adversarial agents to "red team" the code and find those sorts of subtle errors is quite exciting, but I'm not going to bet the farm on it yet.
You are clearly over analyzing a poll that doesn’t hold up to much scrutiny. 44-41 (85% total). There’s simply no credible way that percentage of people actually have an opinion on this more niche topic.
On #1, I think it depends on what you call fully functional project. If you mean a consumer-facing piece of software for thousands of people, sure, maybe — but if you mean perfectly functional ways to speed up business processes or personal productivity, this is totally within the realm
On #2, I agree. See my reference to Kapoor and Narayan for why it's not even applicable to coding yet, which is perhaps most at risk. Also see a few earlier pieces I did on the jobs story. The point is not that it's happening — it's that the perception that it's happening is enough to gin up substantial support for a ban
[In our poll, we asked responders how they use AI. A majority said they used AI for looking up information (55%), while just 8% used it for coding, despite the immense range of self-starting projects it could facilitate.]
Hey! Listen, as one of the 55% I use AI for looking up information because information is what I need. I don't use it to code because there's nothing I want to code. There are no websites I need to build. I am not trying to run a business. There aren't places in my life where I am going, "Oh if only I had a piece of software coded for my specific needs." That's not to say that there couldn't be places where it would be great to have software coded for my specific needs, but I don't know what those places are.
I'm with you, I think that's actually what's super interesting about the result — we tend to think about coding as designed for weird software-y things, but there's actually a ton of stuff amenable to a coded solution.
There's that saying, "To the man with a hammer, everything looks like a nail."
I've sort of taken the approach lately that it's good to scan the world this way — what issues am I having that a coding project could solve. It's fun, and I think it gives us all skills we'll need in the near future.
Anyway, I don't look at as judging people who don't code. I didn't until recently. I look at it as as surprising think that there's this super easy way to start and were not even close to a bunch of people messing around with it
The header of this article is very funny, but I found the analysis unpersuasive. The poll results were surprising but muddled, and I suspect that a non-negligible portion of the respondents may have misunderstood the poll question, which was confusingly worded (I had to read it twice, and can imagine many people may have selected "oppose" to mean the opposite, as in "I oppose human workers being replaced by AI in K12 education).
It seems notable to me that there were in fact three areas where a majority opposed the ban, and they were all the driving ones: public transportation, taxis, and trucking.
> This makes sense because younger people, earlier in their careers, have more of a stake in any given ban.
This is a disappointingly shallow analysis.
Historically, it was often older cohorts who more opposed technological change, whereas younger people were seen as easier to retrain.
Today's case has specific interesting features:
1. Unemployment is specifically increasing among young people. The popular theory is that AI is (currently) good enough to replace juniors but not seniors. Of course, this may not actually be true. Another likely explanation is executives combining typical cost-cutting measures with AI-oriented rhetoric to juice their stock prices (or just mindlessly following the herd).
2. Anti-AI sentiment overall is currently more common on the left, and the younger cohorts skew in that direction.
"As of right now, however, not many people are using AI to code, despite the fact that it requires zero prior coding knowledge to produce fully functional projects. With AI, if you want to start a business, you can not only create your website with AI, but you can now run all your business processes — from expense tracking to payroll — with a set of prepackaged commands."
I'm a non-software engineer who has been using coding agents for months now both at work and for personal projects and gets a lot of value out of them, and I don't agree with this. Yes, you can create little programs and scripts reliably with AI, but I would not trust it to write enterprise-level code, and I especially wouldn't just ask Claude to run crucial parts of a small business as your link suggests. I would not recommend small business owners rely on AI to that degree.
Why not? For many reasons:
1. If something goes wrong, you as the business owner are liable rather than Anthropic.
2. Enterprise code is reliable, efficient, secure, and built for myriad edge cases in a way that a vibe coded app is very unlikely to be (and you wouldn't be qualified to evaluate whether it is or not).
3. Hallucinations have become a lot less common, but they do still happen, and especially for process like payroll where you need to be right 100% of the time, they're still not reliable enough.
I've been working with Claude inside VSCode a bit, over the last few months, and while it's extremely useful, sometimes it makes boneheaded errors. Worse, sometimes it makes more subtle mistakes that require careful review to catch and correct.
The possibility that Mythos will be able to set up adversarial agents to "red team" the code and find those sorts of subtle errors is quite exciting, but I'm not going to bet the farm on it yet.
> enterprise code is reliable, efficient, secure...
Lol. Lmao even
You are clearly over analyzing a poll that doesn’t hold up to much scrutiny. 44-41 (85% total). There’s simply no credible way that percentage of people actually have an opinion on this more niche topic.
1. "despite the fact that it requires zero prior coding knowledge to produce fully functional projects."
Laughably untrue, and even the most enthusiastic vibe coding enthusiasts will tell you so
2. There is no evidence that a meaningful number of American jobs have been taken by AI to date.
On #1, I think it depends on what you call fully functional project. If you mean a consumer-facing piece of software for thousands of people, sure, maybe — but if you mean perfectly functional ways to speed up business processes or personal productivity, this is totally within the realm
On #2, I agree. See my reference to Kapoor and Narayan for why it's not even applicable to coding yet, which is perhaps most at risk. Also see a few earlier pieces I did on the jobs story. The point is not that it's happening — it's that the perception that it's happening is enough to gin up substantial support for a ban
[In our poll, we asked responders how they use AI. A majority said they used AI for looking up information (55%), while just 8% used it for coding, despite the immense range of self-starting projects it could facilitate.]
Hey! Listen, as one of the 55% I use AI for looking up information because information is what I need. I don't use it to code because there's nothing I want to code. There are no websites I need to build. I am not trying to run a business. There aren't places in my life where I am going, "Oh if only I had a piece of software coded for my specific needs." That's not to say that there couldn't be places where it would be great to have software coded for my specific needs, but I don't know what those places are.
I'm with you, I think that's actually what's super interesting about the result — we tend to think about coding as designed for weird software-y things, but there's actually a ton of stuff amenable to a coded solution.
There's that saying, "To the man with a hammer, everything looks like a nail."
I've sort of taken the approach lately that it's good to scan the world this way — what issues am I having that a coding project could solve. It's fun, and I think it gives us all skills we'll need in the near future.
Anyway, I don't look at as judging people who don't code. I didn't until recently. I look at it as as surprising think that there's this super easy way to start and were not even close to a bunch of people messing around with it
It’s worth thinking about those places
The header of this article is very funny, but I found the analysis unpersuasive. The poll results were surprising but muddled, and I suspect that a non-negligible portion of the respondents may have misunderstood the poll question, which was confusingly worded (I had to read it twice, and can imagine many people may have selected "oppose" to mean the opposite, as in "I oppose human workers being replaced by AI in K12 education).
It seems notable to me that there were in fact three areas where a majority opposed the ban, and they were all the driving ones: public transportation, taxis, and trucking.
> This makes sense because younger people, earlier in their careers, have more of a stake in any given ban.
This is a disappointingly shallow analysis.
Historically, it was often older cohorts who more opposed technological change, whereas younger people were seen as easier to retrain.
Today's case has specific interesting features:
1. Unemployment is specifically increasing among young people. The popular theory is that AI is (currently) good enough to replace juniors but not seniors. Of course, this may not actually be true. Another likely explanation is executives combining typical cost-cutting measures with AI-oriented rhetoric to juice their stock prices (or just mindlessly following the herd).
2. Anti-AI sentiment overall is currently more common on the left, and the younger cohorts skew in that direction.
The children yearn for the coding mines