I remain a bit of an AI skeptic. I work in the tech industry in an engineering-adjacent role (anatlyics) and I do not see that AI has improved the pace of software engineering.
It is true that AI produces *code* very quickly, but producing code is only one part of software engineering. I think it's entirely likely that even the currently magical-seeming AI coding products are actually a net negative when all is said and done.
My thinking about the AI hype cycle is somewhat informed by Dan Davies' recent "Snobby about Excel" essay (here: https://backofmind.substack.com/p/snobby-about-excel). AI coding tools seem genuinely magical, especially to non-technical people. They probably represent a significant advance in terms of the abilities of end-user computing. It's not clear how disruptive they'll actually be to more technical engineering work, or to deeper knowledge work more generally.
I also think that some of the AI hype represents a longing for *something* to happen. The material conditions of the US circa 2026 feel quite degraded. Furthermore, the tech industry really hasn't produced much of note since what... the iPhone? We have garbage social media platforms, exploitative gig work platforms, and brain-numbing entertainment platforms. I think there's a real disappointment with what the 00's promise of technology has wrought, and that combines with dissatisfaction with the status quo to produce this longing for something, anything to be different in a big way.
I wonder what happens if after all of the hype, these AI agents just end up amounting to a modestly upgraded version of Excel.
As a programmer, I feel quite similar to this, and have for a while now. People have been promising revolutionary tools to make programming obsolete for decades, and the industry is very trend-driven. We have very little information about the long-term engineering sustainability of AI-written code, but if I had to guess I'd say it's probably not great. It might be neat for one-off little scripts or for non-programmers to use (kinda like Excel), but I'm skeptical it'll have as big an effect on actual software engineering as people seem to think.
edit: this is kind of a tangent, sorry, but I do also think... for however capable the technology may or may not be, it still remains to be seen if anyone can run a profitable business selling access to these state-of-the-art LLMs. I'm still kind of at a point of "I'll believe it when I see it" with regards to the long-term future of AI -- if Anthropic goes bankrupt and can't afford electricity, it doesn't really matter how good their very-energetically-expensive AI model was when their power usage was being subsidized. It's possible only a worse AI model would be profitable to run, but such a model would definitely be less revolutionary.
I think a pretty significant reason that people are grasping for the "pandemic" metaphor is precisely what you're saying—it's the last time Something Happened. I suppose part of what I'm trying to say in this post is that I'm not sure that A.I. deployment/expansion/advance is actually ever going to feel like Something Happening in the way that I think a lot of boosters would like!
I think "A.I. is basically the next spreadsheet revolution" is not such a bad comp, either—and not at all an insignificant thing.
I think you’re right in that the Something will never actually Happen. AI is going to end up being a tool that somewhat changes work patterns but ultimately not in a way that represents the discontinuity people are seeking.
Assume its significance is similar to the internet. It will certainly change society but those changes will be incremental and perhaps not as materially significant as people like to believe (the Bob Gordon story, basically).
Especially since unlike a lot of .com era tech this stuff isn’t being made by scrappy startups but rather by huge firms that are already embedded in the power and social relations of mature tech, I expect to be very disappointed.
Great bottomline question, Jeff! If nothing else, a lot of people will be out a LOT of money if we just end up with easier-to-use-Excel. But I ain't mad about it.
The deindustrialization comparison feels the most apt to me as someone in tech. A lot of factories in the US closed but some stayed open and just went from having 1,000 people on staff to having 100 people who operate robots and oversee automated processes. No one knows right now how to be one of those 100 people and the discourse is finally catching up to this widespread anxiety.
100. And this is combined with all the offshoring the tech companies are doing too. Working in tech today feels like working in the auto industry in the 80s. It was a plush comfortable reliable job... now it's being moved overseas or automated by AI. And how much of that "automation by AI" is just an excuse to lay people off?
Yeah, as the John Herrman piece I link to suggests I think this is a driving emotion behind the virality of that Twitter essay—lots of programmers out there waiting for everyone else (even other programmers) to feel the same sense of doom they do.
“When people ask, ‘When will this end?,’ they are asking about the social ending" is what I'm going to say to Ari next time she complains that she's hearing about A.I. too much
Can someone explain the Covid comparison to me because I don’t get it at all?
I've been giving talks about AI and the future of work since 2013. There have been thousands of articles, books, tv shows and other cultural artifacts discussing this topic for decades. Companies have used AI as an excuse for firing people, etc.
It’s the exact opposite of Covid, where only some experts (and Bill Gates) have been warning about the danger of a pandemic.
Many fewer people would have read the article by Shumer if it had been the pendant of someone warning about Covid in January 2020. Instead, it got so much attention precisely because it manifests the current dominant future imaginary.
This seems a lot like 2000 when the the CFO of the largest US construction company convinced the family to bet it's entire free cash flow on a billion dollar project with Web Van (being a private company it didn’t have to disclose this) to build the infrastructure to deliver groceries to every house in the US. Great idea, but 20 years ahead of the time when the idea would be successfully implemented by paying gig workers less then minimum wage.
Web Van didn’t survive long and the billions invested in it were lost. That construction company survived its cash crunch by firing the CFO, opening its books to banks for the first time to get loans, and putting in place strict project approvals and cash flow management processes. Oh, and it just so happened the LNG export business exploded in 2001, which tripled revenues over the next 10 years. The current CEO was a teenage summer intern in the IT dept attending Stanford at the time. Nobody has ever mentioned Web Van in 20 years.
Its just impossible to forecast how over a trillion dollars capital investment of virtually the entire free cash flow of FANG and all the debt load for Oracle is going to work out. I suspect the Web Van experience has some lesson in it.
The dot-com-bubble was not about "anything is possible right now enthusiasm". It was about "we're going to make a ton of money at the IPO" enthusiasm. There's a big difference.
The right wing discourse loves the “we don’t need software engineers anymore” after the “everyone should go to college for coding” that they believed was the left yelling at them 15 years ago. I think that partially explains the Twitter attention.
“Admittedly, the widget will only save me 10 or so minutes of busy work every week, but suddenly, a whole host of accumulated but untouched wouldn’t-that-be-nice-to-have ideas for widgets and apps and pages and features has opened itself up to me.”
This captures something I find puzzling—it makes me feel like I inhabit a different universe than many other technology people.
I’ve worked as a software engineer at a global telecommunications company for 25 years. I even run my own email server on OpenBSD. Yet coding and technology aren’t my hobby, they’re my job. When I’m done writing software for the day, I step away from computers. I’ve never accumulated a backlog of little coding projects or ideas for apps and widgets. I don’t need any of that in my life, and I genuinely don’t understand why so many others seem to.
What I actually need is to rewire the electrical system on my sailboat, fix one of my toilets, or repair the small plumbing leak in my basement.
Yeah, I think that's also a pretty common reaction. I wanted to write in this post (but didn't have a good spot) a theory I have that some level of the outsized awed reaction you see on X is because X probably over-samples people who have the kind of jobs where "instantly create busy-work-eliminating apps" is a really attractive proposition—freelancers, small-business owners, independent journalists/contractors/consultants, and even employed people with really highly autonomous roles. I can imagine lots of roles where that particular type of agentic A.I. use just isn't that interesting.
@MaxRead I read your headline, and because I'm "too online," in some areas of the Zinternet, immediately thought I missed the beginning of mass layoffs at OpenAI or more at Salesforce, or other tech companies that have wholly embraced the "AGI is the future" message, only to realize they need more than just a message to *checks notes* actually make money.
And it's related to the great up-staffing-then-down-staffing that the Videogame industry saw in the last 3-4 years, immediately following the Pandemic. First, the entire industry saw the HUUUUGE surge in gaming (players, viewers, interest, etc) and thought "look at all this pent up demand! We MUST MEET IT!" +30,000 Jobs! Too late they realize, "Whoops, it was because no one could really socialize and was an artificial upturn. Looks like we SHOULDNT have doubled staff sizes.🤷♀️" -30,000 jobs in one year (low ball estimate).
Quasi-related, why in the world have none of these AGI startups/companies investing heavily in them ALSO started investing heavily in green energy? I mean, maybe I'm a moron, but if you've got a data center, putting solar panels right on top of massive flat space seems a no brainer. Or building a few massive wind turbines right there...boom, lots of power. And wouldn't you be wanting to get this investment to keep going, instead of being told to bring back coal?! Like, I'm sorry, but are the computers supposed to be steam-powered? Didn't Terry Pratchett ALREADY make fun of this 15 or 20 years ago? Steampunk is an aesthetic, not the way we actually power our economy.
I liked this. These days, whenever I read a very contrarian take (and I'm talking the extremes of someone still using the term "stochastic parrots" or pointing to NFT scams), I wonder to myself how long it will be before this technology will be normal enough that it's just part of the background like GPS. That is to say, how long before the fact that your phone can take a picture of an empty room and redecorate on the fly is just expected behavior rather than the basis for a Google commercial?
I think that moment is coming, probably faster than most people think, regardless of any changes to the labor market.
I remain a bit of an AI skeptic. I work in the tech industry in an engineering-adjacent role (anatlyics) and I do not see that AI has improved the pace of software engineering.
It is true that AI produces *code* very quickly, but producing code is only one part of software engineering. I think it's entirely likely that even the currently magical-seeming AI coding products are actually a net negative when all is said and done.
My thinking about the AI hype cycle is somewhat informed by Dan Davies' recent "Snobby about Excel" essay (here: https://backofmind.substack.com/p/snobby-about-excel). AI coding tools seem genuinely magical, especially to non-technical people. They probably represent a significant advance in terms of the abilities of end-user computing. It's not clear how disruptive they'll actually be to more technical engineering work, or to deeper knowledge work more generally.
I also think that some of the AI hype represents a longing for *something* to happen. The material conditions of the US circa 2026 feel quite degraded. Furthermore, the tech industry really hasn't produced much of note since what... the iPhone? We have garbage social media platforms, exploitative gig work platforms, and brain-numbing entertainment platforms. I think there's a real disappointment with what the 00's promise of technology has wrought, and that combines with dissatisfaction with the status quo to produce this longing for something, anything to be different in a big way.
I wonder what happens if after all of the hype, these AI agents just end up amounting to a modestly upgraded version of Excel.
As a programmer, I feel quite similar to this, and have for a while now. People have been promising revolutionary tools to make programming obsolete for decades, and the industry is very trend-driven. We have very little information about the long-term engineering sustainability of AI-written code, but if I had to guess I'd say it's probably not great. It might be neat for one-off little scripts or for non-programmers to use (kinda like Excel), but I'm skeptical it'll have as big an effect on actual software engineering as people seem to think.
edit: this is kind of a tangent, sorry, but I do also think... for however capable the technology may or may not be, it still remains to be seen if anyone can run a profitable business selling access to these state-of-the-art LLMs. I'm still kind of at a point of "I'll believe it when I see it" with regards to the long-term future of AI -- if Anthropic goes bankrupt and can't afford electricity, it doesn't really matter how good their very-energetically-expensive AI model was when their power usage was being subsidized. It's possible only a worse AI model would be profitable to run, but such a model would definitely be less revolutionary.
I think a pretty significant reason that people are grasping for the "pandemic" metaphor is precisely what you're saying—it's the last time Something Happened. I suppose part of what I'm trying to say in this post is that I'm not sure that A.I. deployment/expansion/advance is actually ever going to feel like Something Happening in the way that I think a lot of boosters would like!
I think "A.I. is basically the next spreadsheet revolution" is not such a bad comp, either—and not at all an insignificant thing.
I think you’re right in that the Something will never actually Happen. AI is going to end up being a tool that somewhat changes work patterns but ultimately not in a way that represents the discontinuity people are seeking.
Assume its significance is similar to the internet. It will certainly change society but those changes will be incremental and perhaps not as materially significant as people like to believe (the Bob Gordon story, basically).
Especially since unlike a lot of .com era tech this stuff isn’t being made by scrappy startups but rather by huge firms that are already embedded in the power and social relations of mature tech, I expect to be very disappointed.
Great bottomline question, Jeff! If nothing else, a lot of people will be out a LOT of money if we just end up with easier-to-use-Excel. But I ain't mad about it.
The deindustrialization comparison feels the most apt to me as someone in tech. A lot of factories in the US closed but some stayed open and just went from having 1,000 people on staff to having 100 people who operate robots and oversee automated processes. No one knows right now how to be one of those 100 people and the discourse is finally catching up to this widespread anxiety.
100. And this is combined with all the offshoring the tech companies are doing too. Working in tech today feels like working in the auto industry in the 80s. It was a plush comfortable reliable job... now it's being moved overseas or automated by AI. And how much of that "automation by AI" is just an excuse to lay people off?
Yeah, as the John Herrman piece I link to suggests I think this is a driving emotion behind the virality of that Twitter essay—lots of programmers out there waiting for everyone else (even other programmers) to feel the same sense of doom they do.
Keeping with the pandemic metaphor, your closing reminded me of this prescient NY Times essay from May 2020 on how pandemics end.
https://www.nytimes.com/2020/05/10/health/coronavirus-plague-pandemic-history.html
More or less: never, but we eventually get tired of talking about it.
“When people ask, ‘When will this end?,’ they are asking about the social ending" is what I'm going to say to Ari next time she complains that she's hearing about A.I. too much
Can someone explain the Covid comparison to me because I don’t get it at all?
I've been giving talks about AI and the future of work since 2013. There have been thousands of articles, books, tv shows and other cultural artifacts discussing this topic for decades. Companies have used AI as an excuse for firing people, etc.
It’s the exact opposite of Covid, where only some experts (and Bill Gates) have been warning about the danger of a pandemic.
Many fewer people would have read the article by Shumer if it had been the pendant of someone warning about Covid in January 2020. Instead, it got so much attention precisely because it manifests the current dominant future imaginary.
I appreciated this, after what has felt like a particularly dramatic week in tech AI land.
This seems a lot like 2000 when the the CFO of the largest US construction company convinced the family to bet it's entire free cash flow on a billion dollar project with Web Van (being a private company it didn’t have to disclose this) to build the infrastructure to deliver groceries to every house in the US. Great idea, but 20 years ahead of the time when the idea would be successfully implemented by paying gig workers less then minimum wage.
Web Van didn’t survive long and the billions invested in it were lost. That construction company survived its cash crunch by firing the CFO, opening its books to banks for the first time to get loans, and putting in place strict project approvals and cash flow management processes. Oh, and it just so happened the LNG export business exploded in 2001, which tripled revenues over the next 10 years. The current CEO was a teenage summer intern in the IT dept attending Stanford at the time. Nobody has ever mentioned Web Van in 20 years.
Its just impossible to forecast how over a trillion dollars capital investment of virtually the entire free cash flow of FANG and all the debt load for Oracle is going to work out. I suspect the Web Van experience has some lesson in it.
Feel like Webvan (and also Kozmo) are emblematic of dot-com-bubble 'we live in the future! anything is possible *right now!*' enthusiasm.
The dot-com-bubble was not about "anything is possible right now enthusiasm". It was about "we're going to make a ton of money at the IPO" enthusiasm. There's a big difference.
"If you want a picture of the future, imagine an AI slopping on a human face -- forever."
The right wing discourse loves the “we don’t need software engineers anymore” after the “everyone should go to college for coding” that they believed was the left yelling at them 15 years ago. I think that partially explains the Twitter attention.
“Admittedly, the widget will only save me 10 or so minutes of busy work every week, but suddenly, a whole host of accumulated but untouched wouldn’t-that-be-nice-to-have ideas for widgets and apps and pages and features has opened itself up to me.”
This captures something I find puzzling—it makes me feel like I inhabit a different universe than many other technology people.
I’ve worked as a software engineer at a global telecommunications company for 25 years. I even run my own email server on OpenBSD. Yet coding and technology aren’t my hobby, they’re my job. When I’m done writing software for the day, I step away from computers. I’ve never accumulated a backlog of little coding projects or ideas for apps and widgets. I don’t need any of that in my life, and I genuinely don’t understand why so many others seem to.
What I actually need is to rewire the electrical system on my sailboat, fix one of my toilets, or repair the small plumbing leak in my basement.
Yeah, I think that's also a pretty common reaction. I wanted to write in this post (but didn't have a good spot) a theory I have that some level of the outsized awed reaction you see on X is because X probably over-samples people who have the kind of jobs where "instantly create busy-work-eliminating apps" is a really attractive proposition—freelancers, small-business owners, independent journalists/contractors/consultants, and even employed people with really highly autonomous roles. I can imagine lots of roles where that particular type of agentic A.I. use just isn't that interesting.
@MaxRead I read your headline, and because I'm "too online," in some areas of the Zinternet, immediately thought I missed the beginning of mass layoffs at OpenAI or more at Salesforce, or other tech companies that have wholly embraced the "AGI is the future" message, only to realize they need more than just a message to *checks notes* actually make money.
And it's related to the great up-staffing-then-down-staffing that the Videogame industry saw in the last 3-4 years, immediately following the Pandemic. First, the entire industry saw the HUUUUGE surge in gaming (players, viewers, interest, etc) and thought "look at all this pent up demand! We MUST MEET IT!" +30,000 Jobs! Too late they realize, "Whoops, it was because no one could really socialize and was an artificial upturn. Looks like we SHOULDNT have doubled staff sizes.🤷♀️" -30,000 jobs in one year (low ball estimate).
Quasi-related, why in the world have none of these AGI startups/companies investing heavily in them ALSO started investing heavily in green energy? I mean, maybe I'm a moron, but if you've got a data center, putting solar panels right on top of massive flat space seems a no brainer. Or building a few massive wind turbines right there...boom, lots of power. And wouldn't you be wanting to get this investment to keep going, instead of being told to bring back coal?! Like, I'm sorry, but are the computers supposed to be steam-powered? Didn't Terry Pratchett ALREADY make fun of this 15 or 20 years ago? Steampunk is an aesthetic, not the way we actually power our economy.
I liked this. These days, whenever I read a very contrarian take (and I'm talking the extremes of someone still using the term "stochastic parrots" or pointing to NFT scams), I wonder to myself how long it will be before this technology will be normal enough that it's just part of the background like GPS. That is to say, how long before the fact that your phone can take a picture of an empty room and redecorate on the fly is just expected behavior rather than the basis for a Google commercial?
I think that moment is coming, probably faster than most people think, regardless of any changes to the labor market.