3 ms·

> The amount of software I can create is now mostly limited by ... hard thinking. I'm a product manager and I was talking to my dev lead yesterday about this v

by gcanyon 10mo ago

> The amount of software I can create is now mostly limited by ... hard thinking. I'm a product manager and I was talking to my dev lead yesterday about this very thing. If PMs are like headlights on a car and devs are like the engine, then we're going from cars that max at 80mph to cars that push past 600mph, and we're headed toward much faster than that. The headlights need to extend much further into the future to be able to keep the car from repeatedly running into things. That's the challenge we face. To paraphrase Ian Malcolm, we need to think beyond what we can build to consider more deeply what we should build.

christophilus 10mo ago
The dev role will become a hybrid of product management and micro-manager-style middle management. It’s not a role I particularly want to inhabit, however, so as an individual contributor, I don’t know what I want to do next in my career. I’ve intentionally avoided the management track for 25 years.
svantana 10mo ago
> 80mph to cars that push past 600mph I have yet to see evidence that this is really the case. Already 15 years ago, people were creating impressive software over the course of a hackday, by glueing open source repos together in a high level language. Now that process has been sped up even more, but does it matter that much if the prototype takes 4 or 24 hours to make? The real value is in well-thought-out, highly polished apps, and AFAICT those still take person-years to complete.
- elevation 10mo ago
  The REAL speed up comes to efforts that are already well-designed, but require lots of human busy work. I've personally seen multi-day human efforts reduced to a 15-minute session with an LLM. In a way, LLMs are reducing implementation costs to the kolmolgorov complexity -- you can get what you prompt for, but you have to remember to prompt for everything you want to get -- which comes easiest if you already took time to consider the design.
Zababa 10mo ago
At the risk of extending the metaphor to the point where it is useless, since you go that much faster you can figure out where to go by bumping into things. For example, iterating on an interface by swiping left or right on stuff that you like, first for the general structure of the interface, and then on the details. An example would be, when taking photos these days since we have digital cameras I tend to take a lot of them, and then cull afterwards. Vs with film where I used to take way less. So now you have the whole culling phase where you need different tools, different skills, and also where you risk not putting the same amount of effort in photos in the first place and ending up with way more photos, none of them good (the so-called "slop"). Still, it is possible to take more and better pictures than before with digital cameras.
- gcanyon 10mo ago
  Coincidence, my stepfather was a wedding photographer for about six years. His saying (this was definitely in the era of physical film) was: "Great photographers don't always take great photographs, they just don't show anyone their mistakes." To your point: I think we're a fair bit away from Roomba-style (run in a single direction until you bump into something, then back up, turn, and repeat) development, but it doesn't seem impossible.
delusional 10mo ago
"If my grandmother had wheels she would have been a bike" Development is nothing like driving a car. It makes no sense to liken it to driving a car. There's no set route to go, no road to follow, nor a person solely driving it.

by gcanyon 10mo ago

christophilus 10mo ago

The dev role will become a hybrid of product management and micro-manager-style middle management. It’s not a role I particularly want to inhabit, however, so as an individual contributor, I don’t know what I want to do next in my career. I’ve intentionally avoided the management track for 25 years.

svantana 10mo ago

> 80mph to cars that push past 600mph I have yet to see evidence that this is really the case. Already 15 years ago, people were creating impressive software over the course of a hackday, by glueing open source repos together in a high level language. Now that process has been sped up even more, but does it matter that much if the prototype takes 4 or 24 hours to make? The real value is in well-thought-out, highly polished apps, and AFAICT those still take person-years to complete.

elevation 10mo ago
The REAL speed up comes to efforts that are already well-designed, but require lots of human busy work. I've personally seen multi-day human efforts reduced to a 15-minute session with an LLM. In a way, LLMs are reducing implementation costs to the kolmolgorov complexity -- you can get what you prompt for, but you have to remember to prompt for everything you want to get -- which comes easiest if you already took time to consider the design.

Zababa 10mo ago

At the risk of extending the metaphor to the point where it is useless, since you go that much faster you can figure out where to go by bumping into things. For example, iterating on an interface by swiping left or right on stuff that you like, first for the general structure of the interface, and then on the details. An example would be, when taking photos these days since we have digital cameras I tend to take a lot of them, and then cull afterwards. Vs with film where I used to take way less. So now you have the whole culling phase where you need different tools, different skills, and also where you risk not putting the same amount of effort in photos in the first place and ending up with way more photos, none of them good (the so-called "slop"). Still, it is possible to take more and better pictures than before with digital cameras.

gcanyon 10mo ago
Coincidence, my stepfather was a wedding photographer for about six years. His saying (this was definitely in the era of physical film) was: "Great photographers don't always take great photographs, they just don't show anyone their mistakes." To your point: I think we're a fair bit away from Roomba-style (run in a single direction until you bump into something, then back up, turn, and repeat) development, but it doesn't seem impossible.

delusional 10mo ago

"If my grandmother had wheels she would have been a bike" Development is nothing like driving a car. It makes no sense to liken it to driving a car. There's no set route to go, no road to follow, nor a person solely driving it.