@itsolelehmann: this was probably my favorite read on AI this year. kudos to Dan for writing it his core point: more AI = more work for…

this was probably my favorite read on AI this year. kudos to Dan for writing it

his core point:

more AI = more work for humans, not less

this seems to be universally true: AI is helping everyone get more done, but me and everyone i know is working more than ever, not less

here’s why that’s happening:

1. AI is really good at the stuff humans used to be good at. writing, coding, design, research. it can do all of it now, fast and cheap

2. so everyone starts using it. the person in customer support is shipping code. the engineer is writing blog posts. the marketer is making videos. way more stuff gets made

3. but everyone is using the same AI. so the stuff it makes all starts to look the same

4. when everything looks the same, people get bored of it. they can smell it a mile away. it becomes worthless

5. so now what’s valuable is the opposite. stuff that doesn’t look like everything else. stuff that’s actually good. stuff that fits the exact moment

6. and the only one who can do that is a human who knows what they’re doing. someone who can look at what the AI made and say “this part is good, this part is bad, do it again like this”

7. that human now has way more work to do than before. because the AI is making 10x more stuff, and all of it needs a human to make it actually good

you can see this play out every time a new AI model drops.

people panic that the jobs are gone.

but the same thing happens every time. the AI gets really good at the part of the job that used to be hard.

that part becomes easy and cheap. and the humans who used to do it move up to a new harder part the AI can’t do yet

example: writing a decent blog post used to be a skill. now AI can do it in 10 seconds. so the human job moved up a level.

now the skill is knowing which blog post to write, who it’s for, and whether the one the AI wrote is actually any good

the bar keeps moving up. the humans keep moving with it

even when AGI shows up, someone still has to tell it what to work on and check if the work is any good

the easiest way to see why is to think about a toddler for a second.

a toddler can’t write code, can’t summarize a spreadsheet, can’t pass a single benchmark. the AI beats him at everything

but the toddler wants stuff.

he wants to poke the balloon with a fork. he wants to see if you’ll laugh when he throws his food. he comes up with his own little experiments all day long

the AI doesn’t want anything. it just sits there until someone tells it what to do

that gap doesn’t close.

the AI can do the work. but a human still has to be the one who wants something done in the first place, and who decides if what came back is actually what they wanted

at Every (Dan’s company), they’ve automated everything they possibly can.

and they have more human work to do than ever. that’s the proof

the people who win this are the ones who get good at directing the AI.

telling it what to work on checking if the work is any good and pointing it at problems nobody has figured out how to solve yet

that’s the new skill.

not doing the work yourself, but rather knowing what work is worth doing and whether the AI actually did it well

read the full piece, it’s worth it

Dan Shipper 📧 (@danshipper): We’ve automated every single thing we can @every with AI agents.

And yet there’s way more human work to do than ever. We’ve gone from 4 -> 30 human employees since GPT-3.

I wrote a report on the structural reasons: how AI makes expert competence cheap, why that drives up demand