This talk of 10x productivity makes me blink a bit.
-
This talk of 10x productivity makes me blink a bit.
I get paid for being a developer, but I'm a developer off in the Starts-with-X space; XML, XSD, XPath, XSLT, XQuery, that stuff. So probably not representative.
I have a short feedback loop; there's a small group of specific people who need to be happy with what I'm doing. I'm in meetings with them.
"How fast can I write code?" is rarely a bottleneck. "Have we got enough test coverage?", sometimes. "What should this do?", often.
Back in the day -- say, 1980 - 2000 when I did software -- the hardest question to get answered was asking the client, "What are you REALLY trying to accomplish?"
-
Back in the day -- say, 1980 - 2000 when I did software -- the hardest question to get answered was asking the client, "What are you REALLY trying to accomplish?"
@weekend_editor Still is!
Arguably worse, because the sense of possibility has expanded while experiencing some loss of precision.
-
Effective bug fixes remove code, as when the experience of having written the thing allows a change of approach to a more appropriate abstraction. Stuff like that.
It's hard to find any way to understand the claimed productivity increases that doesn't amount to "more lines of code! more! more!" and as a metric you couldn't find even a self-respecting abyss willing to advance that one today.
These people have stared into an abyss and dissipated their intellects beyond the tenuous aether.
@graydon I think 10x programmers are like AI - they're much much faster than the rest of us at creating bugs.
-
@graydon I think 10x programmers are like AI - they're much much faster than the rest of us at creating bugs.
@zeborah People who will tell you they are 10X programmers, yeah.
There are some people who really are much more productive than the norm in a coding domain, but they rarely describe themselves as such.
-
@zeborah People who will tell you they are 10X programmers, yeah.
There are some people who really are much more productive than the norm in a coding domain, but they rarely describe themselves as such.
@graydon Oh absolutely. Also as you note they're probably not writing 10x the code, they're fixing 10x the bugs.
-
@graydon Oh absolutely. Also as you note they're probably not writing 10x the code, they're fixing 10x the bugs.
@zeborah Or better design.
People don't talk about design much, on the whole and by and large, which is a pity, because design is where the amount of inescapable bugs lives. Software with a high inescapable bug threshold will never get all that debugged no matter how much work gets applied to it.
(E.g., trying to parse HTML with regular expressions. This is not capable of being a robust general case solution.)
-
@weekend_editor Still is!
Arguably worse, because the sense of possibility has expanded while experiencing some loss of precision.
I retired in 2020, working then as an applied statistician.
It was *always* the hardest to get people to back down from "Can you calculate this stat for me?" to describing what they were actually trying to do. Almost always, the stat they thought they wanted was not the most appropriate one.
So, pretty similar.
-
Effective bug fixes remove code, as when the experience of having written the thing allows a change of approach to a more appropriate abstraction. Stuff like that.
It's hard to find any way to understand the claimed productivity increases that doesn't amount to "more lines of code! more! more!" and as a metric you couldn't find even a self-respecting abyss willing to advance that one today.
These people have stared into an abyss and dissipated their intellects beyond the tenuous aether.
@graydon there's people as claim that "AI" (why is there no voice recognition software which works with sophisticated punctuation?) helps in the design phase, too, but I don't think it really does. I have a whole collection of articles discussing "AI" and productivity and it doesn't look like it makes a huge difference. They might get a 10 to 20% productivity increase from quick code generation, but I think that's about it.
https://shinycroak.blogspot.com/2025/09/ai-and-productivity.html
-
@graydon there's people as claim that "AI" (why is there no voice recognition software which works with sophisticated punctuation?) helps in the design phase, too, but I don't think it really does. I have a whole collection of articles discussing "AI" and productivity and it doesn't look like it makes a huge difference. They might get a 10 to 20% productivity increase from quick code generation, but I think that's about it.
https://shinycroak.blogspot.com/2025/09/ai-and-productivity.html
@graydon back when they did serious management studies on computing (which was in my youth and I'm an old bird) coding was about 15% of the time on a project. Still, I don't rule out the possibility that this technology might eventually contribute. Even a 10% increase in productivity is impressive, just not the "best thing with sliced bread, going to eliminate programmers forever" which "AI" developers are claiming.
-
@graydon I think 10x programmers are like AI - they're much much faster than the rest of us at creating bugs.
-
@zeborah Or better design.
People don't talk about design much, on the whole and by and large, which is a pity, because design is where the amount of inescapable bugs lives. Software with a high inescapable bug threshold will never get all that debugged no matter how much work gets applied to it.
(E.g., trying to parse HTML with regular expressions. This is not capable of being a robust general case solution.)
-
undefined oblomov@sociale.network shared this topic on