You know what else about AI?
-
You know what else about AI?
People who hype it have no idea about diminishing returns
We have decades of research showing that when people only work on the hard stuff at work and don't get sufficient breaks, they burn out faster
If you're selling AI as a tool that "automates the mindless parts of work" then the worker is only left with the hard parts, and they burn out faster
@jilleduffy The evolution of the "case against AI" has been very interesting to watch.
At first it was dismissive, then it went through a "boiling the oceans" phase, then it was "based on stealing", then people who don't use it are Ethical Heroes, then it got kettled in detail like "Claude code quality experts", then it was "hey everyone look at me, I will never use it", then copy.fail & co.
Now: "you'll work harder".
Yes. But you will also get a lot of previously impossible shit done.
-
@jilleduffy The evolution of the "case against AI" has been very interesting to watch.
At first it was dismissive, then it went through a "boiling the oceans" phase, then it was "based on stealing", then people who don't use it are Ethical Heroes, then it got kettled in detail like "Claude code quality experts", then it was "hey everyone look at me, I will never use it", then copy.fail & co.
Now: "you'll work harder".
Yes. But you will also get a lot of previously impossible shit done.
@hopeless @jilleduffy none of that other stuff has been resolved.
-
Knowledge workers who are highly productive balance the "mindless" i.e., easy or low-focus, work with the hard stuff. We work it into our day so that the low-focus tasks provide us with a break. That break is integral to getting back to the high-focus work
@jilleduffy I would argue that on top of that it increases the amount of high focus work, like proof reading AI output.
-
Knowledge workers who are highly productive balance the "mindless" i.e., easy or low-focus, work with the hard stuff. We work it into our day so that the low-focus tasks provide us with a break. That break is integral to getting back to the high-focus work
@jilleduffy What agentic development will do to human psyche will not be pretty.
Having said that, as a project manager with ADHD + some undiagnosed conditions thrown in, I must concede Gemini has been a life saver for me.
"Whose job is to do X?"
"What is the latest agreement on Y?"
"Did we ever cover this or that eventuality?"
"What does Z mean?"
All with sources, and while it's imprecise enough I don't trust it, I usually have enough to do the thing and confirm with the person.
(cont.)
-
You know what else about AI?
People who hype it have no idea about diminishing returns
We have decades of research showing that when people only work on the hard stuff at work and don't get sufficient breaks, they burn out faster
If you're selling AI as a tool that "automates the mindless parts of work" then the worker is only left with the hard parts, and they burn out faster
@jilleduffy Shining bright, but shorter. By design.
-
@jilleduffy What agentic development will do to human psyche will not be pretty.
Having said that, as a project manager with ADHD + some undiagnosed conditions thrown in, I must concede Gemini has been a life saver for me.
"Whose job is to do X?"
"What is the latest agreement on Y?"
"Did we ever cover this or that eventuality?"
"What does Z mean?"
All with sources, and while it's imprecise enough I don't trust it, I usually have enough to do the thing and confirm with the person.
(cont.)
@jilleduffy While I write all my e-mail, tickets, and related things, what it automates for me is the hard stuff: remembering, orienting among dozens of meetings I attend, keeping track of things, the lot.
All of these have always been killing me, and being able to focus on the problem solving has actually been liberating.
Sure, none of this offsets the environmental damage, costs, endangered livelihoods, or copyright infringement.
Typed on my oppressive device filled with rare metals.
-
You know what else about AI?
People who hype it have no idea about diminishing returns
We have decades of research showing that when people only work on the hard stuff at work and don't get sufficient breaks, they burn out faster
If you're selling AI as a tool that "automates the mindless parts of work" then the worker is only left with the hard parts, and they burn out faster
@jilleduffy interesting I hadn’t thought of that and I’ve been feeling this recently. I’ve been using it to handle low hanging fruit. I only recently started using it more.
-
You know what else about AI?
People who hype it have no idea about diminishing returns
We have decades of research showing that when people only work on the hard stuff at work and don't get sufficient breaks, they burn out faster
If you're selling AI as a tool that "automates the mindless parts of work" then the worker is only left with the hard parts, and they burn out faster
When they say "mindless parts of work", these fossil fuel funded fascist tech bros often mean work done by women, immigrants, POC, & the young.
These are the jobs seen as "pointless" by billionaires.
https://www.ft.com/content/946650d6-f61f-4b98-8bb5-c0020c8a205f
https://archive.is/3FOVChttps://www.businessinsider.com/recent-company-layoffs-laying-off-workers-2025
The GOP's war on women includes a war on their employment like teaching, health care, retail, & clerical work.
https://www.newsweek.com/nursing-not-professional-degree-trump-admin-11079650
https://mashable.com/article/ai-alpha-school-trump-administration
-
@jilleduffy interesting I hadn’t thought of that and I’ve been feeling this recently. I’ve been using it to handle low hanging fruit. I only recently started using it more.
@jilleduffy this in combination with meaningless RTO is killing me apparently.
-
You know what else about AI?
People who hype it have no idea about diminishing returns
We have decades of research showing that when people only work on the hard stuff at work and don't get sufficient breaks, they burn out faster
If you're selling AI as a tool that "automates the mindless parts of work" then the worker is only left with the hard parts, and they burn out faster
@jilleduffy I think there was an article of the MIT that stated something similar, calling for making sure workers take meaningfull breaks without having something run in the background to prevent burnout.
Though they were looking at it from a different perspective, that is people picking up work outside their expertise (they failed to address the problem of the worker being unable to check the correctness on processes they are unfamiliar with wiith, which is a yikes from me) -
@davidhmccoy This is the thing. AI might be fine for looking at large data sets of medical images, for example, and finding commonalities among images that did or did not develop into cancer 5 years later. That is a great use case. There may be other great use cases in a very specific profession, such as software development. But that is not what AI "hype" is. The hype says "if you don't figure out how to use it in ALL of your jobs, you will be left behind."
@jilleduffy @davidhmccoy very much agree with this sentiment. #AIslop where the expertise is removed and people told not to question the #slop output is bizarre at best. Human experience is where cometency is gained, not solely through injesting books or data sources.
Doing pattern matching with experts reviewing the output is a reasonable application. But these still need validation and potentially just open up different approaches rather than provide answers per se.
-
. It is insane how much they are pushing it at my job. This one manager asked a coworker to use AI to shorten 4 sentences into bullet points. Four.
@davidhmccoy @jilleduffy I know people who generate tickets in Jira using AI which results in terrible quality and no one knows what to do, however they still believe it’s valuable to do so and won’t stop.
-
@bignose @jilleduffy I like this infant analogy!
If you don't hang out with the infant doing everyday stuff and are only parachuted in when they're crying, you'll have no idea why they're crying!
Whereas if you're there all the time, you can go straight to "Oh, that's their 'I need to be burped again' face". And can probably even automatically prevent the crying by recognizing the face before the crying starts
-
You know what else about AI?
People who hype it have no idea about diminishing returns
We have decades of research showing that when people only work on the hard stuff at work and don't get sufficient breaks, they burn out faster
If you're selling AI as a tool that "automates the mindless parts of work" then the worker is only left with the hard parts, and they burn out faster
@jilleduffy Agree, good post. People also tend to forget that the "mindless parts" are a great way to introduce juniors or new hires into a project or whatever.
-
You know what else about AI?
People who hype it have no idea about diminishing returns
We have decades of research showing that when people only work on the hard stuff at work and don't get sufficient breaks, they burn out faster
If you're selling AI as a tool that "automates the mindless parts of work" then the worker is only left with the hard parts, and they burn out faster
@jilleduffy yeah I agree, there is also another challenge, big tech giants are well positioned to take complete leverage from modern tech upgrades like MCP and using it as a paid service and doing tasks like web automation on the contrary individuals are restricted by buzz words like this is spam, privacy violations etc. double standards
-
You know what else about AI?
People who hype it have no idea about diminishing returns
We have decades of research showing that when people only work on the hard stuff at work and don't get sufficient breaks, they burn out faster
If you're selling AI as a tool that "automates the mindless parts of work" then the worker is only left with the hard parts, and they burn out faster
-
@davidhmccoy This is the thing. AI might be fine for looking at large data sets of medical images, for example, and finding commonalities among images that did or did not develop into cancer 5 years later. That is a great use case. There may be other great use cases in a very specific profession, such as software development. But that is not what AI "hype" is. The hype says "if you don't figure out how to use it in ALL of your jobs, you will be left behind."
@jilleduffy @davidhmccoy this is not an example of generative AI (LLM), the subset of machine learning we all love to hate.
The medical image application is based on identifying common factors where there is a yes/no discriminant: whether the imaged person went on to develop cancer or not.
Generative AI is just fancy auto complete, it has no idea of truth and whatever it seems to claim when you prompt it is never tested for accuracy during training. -
@davidhmccoy This is the thing. AI might be fine for looking at large data sets of medical images, for example, and finding commonalities among images that did or did not develop into cancer 5 years later. That is a great use case. There may be other great use cases in a very specific profession, such as software development. But that is not what AI "hype" is. The hype says "if you don't figure out how to use it in ALL of your jobs, you will be left behind."
@jilleduffy @davidhmccoy
AI can be great for medical images, but not the LLM kind of AI -
You know what else about AI?
People who hype it have no idea about diminishing returns
We have decades of research showing that when people only work on the hard stuff at work and don't get sufficient breaks, they burn out faster
If you're selling AI as a tool that "automates the mindless parts of work" then the worker is only left with the hard parts, and they burn out faster
@jilleduffy there is an ideal reality where automation is inversely proportional to the amount of work people are doing and wages go up to lower the average weekly working hours.
-
@hopeless @jilleduffy none of that other stuff has been resolved.
@mu @jilleduffy Right, the big sump of negative arguments can't be resolved, at least not quickly.
Some of it had a sell-by-date, like beliefs around AI not working or coding assists dumping training data.
Some of it is wrong, like coding assists "based on stealing" when most github stuff they trawled is MIT or other liberal license that's fine with it.
"Boiling the oceans" does not apply to selfhosted AI run on solar.
Some of it is like horse paste to an antivaxxer: evidence doesn't apply.
