Great. Too bad I practically never see my desktop. Maybe I should just have a desktop I would never use and get into habit of switching to it occasionally to see the picture?
If you right-click on the icon, the windows will move out of the way and show your desktop! (This is doing the same thing as the trackpad show-desktop gesture.)
The distribution of programming talent is likely normal, but what about their output?
The ‘10X programmer’ is relatively common, maybe 1 standard deviation from the median? And you don’t have to get very far to the left of the curve to find people who are 0.1X or -1.0X programmers.
Still a good article! I think this confusion is the smallest part of what he’s trying to say.
They exist, though. So, more like the model that excludes them is broken front and center. Accurate position is most people aren’t 10x’ers or even need to be that I can tell. Team players with consistency are more valuable in the long run. That should be majority with some strong, technical talent sprinkled in.
Is there evidence to support that? As you know, measuring programmer productivity is notoriously difficult, and I haven’t seen any studies to confirm the 10x difference. I agree with @SeanTAllen, it’s more like an instance of the hero myth.
Just think back to school or college where people got the same training. Some seemed natural at the stuff running circles around others for whatever reason, right? And some people score way higher than others on parts of math, CompSci, or IQ tests seemingly not even trying compared to those that put in much effort to only underperform.
People that are super-high performers from the start exist. If they and the others study equally, the gap might shrink or widen but should widen if wanting strong generalists since they’re better at foundational skills or thinking style. I don’t know if the 10 applies (probably not). The concept of gifted folks making easy work of problems most others struggle is something Ive seen a ton of in real life.
Why would they not exist in programming when they exist in everything else would be the more accurate question.
There’s no question that there is difference in intellectual ability. However, I think that it’s highly questionable that it translates into 10x (or whatever-x) differences in productivity.
Partly it’s because only a small portion of programming is about raw intellectual power. A lot of it is just grinding through documentation and integration issues.
Partly it’s because there are complex interactions with other people that constrain a person. Simple example: at one of my jobs people complained a lot about C++ templates because they couldn’t understand them.
Finally, it’s also because the domain a person applies themselves to places other constraints. Can’t get too clever if you have to stay within the confines of a web framework, for example.
I guess there are specific contexts where high productivity could be realised: one person creating something from scratch, or a group of highly talented people who work well together. But those would be exceptional situations, while under the vast majority of circumstances it’s counterproductive to expect or hope for 10x productivity from anyone.
I agree with all of that. I think the multipliers kick in on particular tasks which may or may not produce a net benefit overall given conflicting requirements. Your example of one person being too clever with some code for others to read is an example of that.
I think the 10x is often realized by just understanding the requirements better. For example, maybe the 2 week long solution isn’t really necessary because the 40 lines you can write in the afternoon are all the requirement really required.
There’s no question that there is difference in intellectual ability. However, I think that it’s highly questionable that it translates into 10x (or whatever-x) differences in productivity.
It does not simply depends on how you measure, it depends on what you measure.
And it may be more than “raw intellectual power”. For me it’s usually experience.
As a passionate programmer, I’ve faced more problems and more bugs than my colleagues.
So it often happens that I solve in minutes problems that they have struggled for hours (or even days).
This has two side effects:
managers tends to assign me the worst issues
colleagues tends to ask me when the can’t find a solution
Both of this force me to face more problems and bugs… and so on.
Also such experience make me well versed at architectural design of large applications: I’m usually able to avoid issues and predict with an high precision the time required for a task.
However measuring overall productivity is another thing:
I can literally forget what I did yesterday morning (if it was for a different customer than the one I’m focused now)
at time I’m unable to recognize my own code (with funny effects when I insult or lode it)
when focused, I do not ear people talking at me
I ignore 95% of mails I receive (literally all those with multiple recipients)
being very good at identifying issues during early analysis at times makes some colleague a bit upset
being very good at estimating large projects means that when you compare my estimation with others, mine is usually higher (at times a lot higher) because I see most costs upfront. This usually leads to long and boring meeting where nobody want to take the responsibility to adopt the more expensive solution (apparently) but nobody want to take the risk of alternative ones either…
debating with me tends to become an enormous waste of time…
So when it’s a matter of solving problems by programming, I’m approach the 10x productivity of the myth despite not being particularly intelligent, but overall it really depends on the environment.
This is a good exposition of what a 10x-er might be and jives with my thoughts. Some developers can “do the hard stuff” with little or no guidance. Some developers just can’t, no matter how much coaching and guidance are provided.
For illustration, I base this on one tenure I had as a team lead, where the team worked on some “algorithmically complex” tasks. I had on my team people who were hired on and excelled at the work. I had other developers who struggled. Most got up to an adequate level eventually (6 months or so). One in particular never did. I worked with this person for a year, teaching and guiding, and they just didn’t get it. This particular developer was good at other things though like trouble shooting and interfacing with customers in more of a support role. But the ones who flew kept on flying. They owned it, knew it inside and out.
It’s odd to me that anyone disputes the fact there are more capable developers out there. Sure “productivety” is one measure, and not a good proxy for ability. I personally don’t equate 10x with being productive, that clearly makes no sense. Also I think Fred Brookes Mythical Man Month is the authoritative source on this. I never see it cited in these discussions.
I used to think that, but I’m no longer sure. I’ve seen multiple instances of what I considered absolutely horrible programmers taking the helm, and I fully expected those businesses to fold in a short period of time as a result - but they didn’t! From my point of view, it’s horrible -10x code, but for the business owner, it’s just fine because the business keeps going and features get added. So how do we even measure success or failure, let alone assign quantifiers like 0x?
Oh, I don’t mean code quality, I mean productivity. I know some devs that can work on the same simple task for weeks, miss the deadline, and be move on to a different task that they also don’t finish.
Even if the code they wrote was amazing, they don’t ship enough progress to be of much help.
I’ve encountered it, though it was unrelated to their skill. Depressive episodes, for example, can really block someone. So can burnout, or outside stresses.
Perhaps there are devs who cannot ship code at all, but I’ve only encountered unshipping devs that were in a bad state.
You’re defining programming ability by if a business succeeds though. There are plenty of other instances where programming is not done for the sake of business, though.
That’s true. But my point is that it makes no sense to assign quantifiers to programmer output without actually being able to measure it. In business, you could at least use financials as a proxy measure (obviously not a great one).
Anecdotally, I’m routinely stunned by how productive maintainers of open source frameworks can be. They’re certainly many times more productive than I am. (Maybe that just means I’m a 0.1x programmer, though!)
I’m sure that’s the case sometimes. But are they productive because they have more sense of agency? Because they don’t have to deal with office politics? Because they just really enjoy working on it (as opposed to a day job)? There are so many possible reasons. Makes it hard to establish how and what to measure to determine productivity.
I don’t get why people feel the need to pretend talent is a myth or that 10x programmers are a myth. It’s way more than 10x. I don’t get why so many obviously talented people need to pretend they’re mediocre.
edit: does anyone do this in any other field? Do people deny einstein, mozart, michaelangelo, shakespear, or newton? LeBron James?
Deny what exactly? That LeBron James exists? What is LeBron James a 10x of? Is that Athelete? Basketball player? What is the scale here?
A 10x programmer. I’ve never met one. I know people who are very productive within their area of expertise. I’ve never met someone who I can drop into any area and they are boom 10x more productive and if you say “10x programmer” that’s what you are saying.
This of course presumes that we can manage to define what the scale is. We can’t as an industry define what productive is. Is it lines of code? Story points completed? Features shipped?
Context is a huge factor in productivity. It’s not fair to subtract it out.
I bet you’re a lot more then 10X better then I am at working on Pony… Any metric you want.
I don’t write much C since college, I bet you’re more then 10X better then me in any C project.
You were coding before I was born, and as far as I can tell are near the top of your field.
I’ve been coding most of my life, I’m good at it, the difference is there though.
I know enough to be able to read your code and tell that you’re significantly more skilled then I am.
I bet you’re only a factor of 2 or 3 better at general programming then I am. (Here I am boasting)
In my areas of expertise, I could win some of that back and probably (but I’m not so sure) outperform you.
I’ve only been learning strategies for handling concurrency for 4 years? Every program (certainly every program with a user interface) has to deal with concurrency, your skill in that sub-domain alone could outweigh my familiarity in any environment.
There are tons of programmers out there who can not deal with any amount of concurrency at all in their most familiar environment. There are bugs that they will encounter which they can not possibly fix until they remedy that deficiency, and that’s one piece of a larger puzzle. I know that the right support structure of more experienced engineers (and tooling) can solve this, I don’t think that kind of support is the norm in the industry.
If we could test our programming aptitudes as we popped out of the womb, all bets are off.
This makes me think that “10X programmer” is ill-defined? Maybe we’re not talking about the same thing at all.
No I agree with you. Context is important. As is having a scale. All the conversations I see are “10x exists” and then no accounting for context or defining a scale.
While I’m not very familiar with composers, I can tell you that basketball players (LeBron) can and do have measurements. Newton created fundamental laws and integral theories, Shakespeare’s works continue to be read.
We do acknowledge the groundbreaking work of folks like Ken Ritchie, Ken Iverson, Alan Kay, and other computing pioneers, but I doubt “Alice 10xer” at a tech startup will have her work influence software engineers hundreds of years later, so bar that sort of influence, there are not enough metrics or studies to show that an engineer is 10x more than another in anything.
The ‘10X programmer’ is relatively common, maybe 1 standard deviation from the median? And you don’t have to get very far to the left of the curve to find people who are 0.1X or -1.0X programmers.
So, it’s fairly complicated because people who will be 10X in one context are 1X or even -1X in others. This is why programming has so many tech wars, e.g. about programming languages and methodologies. Everyone’s trying to change the context to one where they are the top performers.
There are also feedback loops in this game. Become known as a high performer, and you get new-code projects where you can achieve 200 LoC per day. Be seen as a “regular” programmer, and you do thankless maintenance where one ticket takes three days.
I’ve been a 10X programmer, and I’ve been less-than-10X. I didn’t regress; the context changed out of my favor. Developers scale badly and most multi-developer projects have a trailblazer and N-1 followers. Even if the talent levels are equal, a power-law distribution of contributions (or perceived contributions) will emerge.
I’m glad you acknowledge that there’s room for a 10X or more then 10X gap in productivity. It surprises me how many people claim that there is no difference in productivity among developers. (Why bother practicing and reading blog posts? It won’t make you better!)
I’m more interested in exactly what it takes to turn a median (1X by definition) developer into an exceptional developer.
I don’t buy the trail-blazer and N-1 followers argument because I’ve witnessed massive success (by any metric) cleaning up the non-functioning, non-requirements meeting (but potentially marketable!) untested messes that an unskilled ‘trailblazer’ leaves in their (slowly moving) wake. Do you think it’s all context or are there other forces at work?
Great. Too bad I practically never see my desktop. Maybe I should just have a desktop I would never use and get into habit of switching to it occasionally to see the picture?
If you right-click on the icon, the windows will move out of the way and show your desktop! (This is doing the same thing as the trackpad show-desktop gesture.)
The distribution of programming talent is likely normal, but what about their output?
The ‘10X programmer’ is relatively common, maybe 1 standard deviation from the median? And you don’t have to get very far to the left of the curve to find people who are 0.1X or -1.0X programmers.
Still a good article! I think this confusion is the smallest part of what he’s trying to say.
That’s an interesting backdoor you tried to open to sneak the 10x programmer back into not being a myth.
They exist, though. So, more like the model that excludes them is broken front and center. Accurate position is most people aren’t 10x’ers or even need to be that I can tell. Team players with consistency are more valuable in the long run. That should be majority with some strong, technical talent sprinkled in.
Is there evidence to support that? As you know, measuring programmer productivity is notoriously difficult, and I haven’t seen any studies to confirm the 10x difference. I agree with @SeanTAllen, it’s more like an instance of the hero myth.
EDIT: here are some interesting comments by a guy who researched the literature on the subject: https://medium.com/make-better-software/the-10x-programmer-and-other-myths-61f3b314ad39
Just think back to school or college where people got the same training. Some seemed natural at the stuff running circles around others for whatever reason, right? And some people score way higher than others on parts of math, CompSci, or IQ tests seemingly not even trying compared to those that put in much effort to only underperform.
People that are super-high performers from the start exist. If they and the others study equally, the gap might shrink or widen but should widen if wanting strong generalists since they’re better at foundational skills or thinking style. I don’t know if the 10 applies (probably not). The concept of gifted folks making easy work of problems most others struggle is something Ive seen a ton of in real life.
Why would they not exist in programming when they exist in everything else would be the more accurate question.
There’s no question that there is difference in intellectual ability. However, I think that it’s highly questionable that it translates into 10x (or whatever-x) differences in productivity.
Partly it’s because only a small portion of programming is about raw intellectual power. A lot of it is just grinding through documentation and integration issues.
Partly it’s because there are complex interactions with other people that constrain a person. Simple example: at one of my jobs people complained a lot about C++ templates because they couldn’t understand them.
Finally, it’s also because the domain a person applies themselves to places other constraints. Can’t get too clever if you have to stay within the confines of a web framework, for example.
I guess there are specific contexts where high productivity could be realised: one person creating something from scratch, or a group of highly talented people who work well together. But those would be exceptional situations, while under the vast majority of circumstances it’s counterproductive to expect or hope for 10x productivity from anyone.
I agree with all of that. I think the multipliers kick in on particular tasks which may or may not produce a net benefit overall given conflicting requirements. Your example of one person being too clever with some code for others to read is an example of that.
I think the 10x is often realized by just understanding the requirements better. For example, maybe the 2 week long solution isn’t really necessary because the 40 lines you can write in the afternoon are all the requirement really required.
It does not simply depends on how you measure, it depends on what you measure.
And it may be more than “raw intellectual power”. For me it’s usually experience.
As a passionate programmer, I’ve faced more problems and more bugs than my colleagues.
So it often happens that I solve in minutes problems that they have struggled for hours (or even days).
This has two side effects:
Both of this force me to face more problems and bugs… and so on.
Also such experience make me well versed at architectural design of large applications: I’m usually able to avoid issues and predict with an high precision the time required for a task.
However measuring overall productivity is another thing:
So when it’s a matter of solving problems by programming, I’m approach the 10x productivity of the myth despite not being particularly intelligent, but overall it really depends on the environment.
This is a good exposition of what a 10x-er might be and jives with my thoughts. Some developers can “do the hard stuff” with little or no guidance. Some developers just can’t, no matter how much coaching and guidance are provided.
For illustration, I base this on one tenure I had as a team lead, where the team worked on some “algorithmically complex” tasks. I had on my team people who were hired on and excelled at the work. I had other developers who struggled. Most got up to an adequate level eventually (6 months or so). One in particular never did. I worked with this person for a year, teaching and guiding, and they just didn’t get it. This particular developer was good at other things though like trouble shooting and interfacing with customers in more of a support role. But the ones who flew kept on flying. They owned it, knew it inside and out.
It’s odd to me that anyone disputes the fact there are more capable developers out there. Sure “productivety” is one measure, and not a good proxy for ability. I personally don’t equate 10x with being productive, that clearly makes no sense. Also I think Fred Brookes Mythical Man Month is the authoritative source on this. I never see it cited in these discussions.
There may not be any 10x developers, but I’m increasingly convinced that there are many 0x (or maybe epsilon-x) developers.
I used to think that, but I’m no longer sure. I’ve seen multiple instances of what I considered absolutely horrible programmers taking the helm, and I fully expected those businesses to fold in a short period of time as a result - but they didn’t! From my point of view, it’s horrible -10x code, but for the business owner, it’s just fine because the business keeps going and features get added. So how do we even measure success or failure, let alone assign quantifiers like 0x?
Oh, I don’t mean code quality, I mean productivity. I know some devs that can work on the same simple task for weeks, miss the deadline, and be move on to a different task that they also don’t finish.
Even if the code they wrote was amazing, they don’t ship enough progress to be of much help.
That’s interesting. I’ve encountered developers who were slow but not ones who would produce nothing at all.
I’ve encountered it, though it was unrelated to their skill. Depressive episodes, for example, can really block someone. So can burnout, or outside stresses.
Perhaps there are devs who cannot ship code at all, but I’ve only encountered unshipping devs that were in a bad state.
You’re defining programming ability by if a business succeeds though. There are plenty of other instances where programming is not done for the sake of business, though.
That’s true. But my point is that it makes no sense to assign quantifiers to programmer output without actually being able to measure it. In business, you could at least use financials as a proxy measure (obviously not a great one).
Anecdotally, I’m routinely stunned by how productive maintainers of open source frameworks can be. They’re certainly many times more productive than I am. (Maybe that just means I’m a 0.1x programmer, though!)
I’m sure that’s the case sometimes. But are they productive because they have more sense of agency? Because they don’t have to deal with office politics? Because they just really enjoy working on it (as opposed to a day job)? There are so many possible reasons. Makes it hard to establish how and what to measure to determine productivity.
I don’t get why people feel the need to pretend talent is a myth or that 10x programmers are a myth. It’s way more than 10x. I don’t get why so many obviously talented people need to pretend they’re mediocre.
edit: does anyone do this in any other field? Do people deny einstein, mozart, michaelangelo, shakespear, or newton? LeBron James?
Deny what exactly? That LeBron James exists? What is LeBron James a 10x of? Is that Athelete? Basketball player? What is the scale here?
A 10x programmer. I’ve never met one. I know people who are very productive within their area of expertise. I’ve never met someone who I can drop into any area and they are boom 10x more productive and if you say “10x programmer” that’s what you are saying.
This of course presumes that we can manage to define what the scale is. We can’t as an industry define what productive is. Is it lines of code? Story points completed? Features shipped?
Context is a huge factor in productivity. It’s not fair to subtract it out.
I bet you’re a lot more then 10X better then I am at working on Pony… Any metric you want. I don’t write much C since college, I bet you’re more then 10X better then me in any C project.
You were coding before I was born, and as far as I can tell are near the top of your field. I’ve been coding most of my life, I’m good at it, the difference is there though. I know enough to be able to read your code and tell that you’re significantly more skilled then I am. I bet you’re only a factor of 2 or 3 better at general programming then I am. (Here I am boasting)
In my areas of expertise, I could win some of that back and probably (but I’m not so sure) outperform you. I’ve only been learning strategies for handling concurrency for 4 years? Every program (certainly every program with a user interface) has to deal with concurrency, your skill in that sub-domain alone could outweigh my familiarity in any environment.
There are tons of programmers out there who can not deal with any amount of concurrency at all in their most familiar environment. There are bugs that they will encounter which they can not possibly fix until they remedy that deficiency, and that’s one piece of a larger puzzle. I know that the right support structure of more experienced engineers (and tooling) can solve this, I don’t think that kind of support is the norm in the industry.
If we could test our programming aptitudes as we popped out of the womb, all bets are off. This makes me think that “10X programmer” is ill-defined? Maybe we’re not talking about the same thing at all.
No I agree with you. Context is important. As is having a scale. All the conversations I see are “10x exists” and then no accounting for context or defining a scale.
While I’m not very familiar with composers, I can tell you that basketball players (LeBron) can and do have measurements. Newton created fundamental laws and integral theories, Shakespeare’s works continue to be read.
We do acknowledge the groundbreaking work of folks like Ken Ritchie, Ken Iverson, Alan Kay, and other computing pioneers, but I doubt “Alice 10xer” at a tech startup will have her work influence software engineers hundreds of years later, so bar that sort of influence, there are not enough metrics or studies to show that an engineer is 10x more than another in anything.
So, it’s fairly complicated because people who will be 10X in one context are 1X or even -1X in others. This is why programming has so many tech wars, e.g. about programming languages and methodologies. Everyone’s trying to change the context to one where they are the top performers.
There are also feedback loops in this game. Become known as a high performer, and you get new-code projects where you can achieve 200 LoC per day. Be seen as a “regular” programmer, and you do thankless maintenance where one ticket takes three days.
I’ve been a 10X programmer, and I’ve been less-than-10X. I didn’t regress; the context changed out of my favor. Developers scale badly and most multi-developer projects have a trailblazer and N-1 followers. Even if the talent levels are equal, a power-law distribution of contributions (or perceived contributions) will emerge.
I’m glad you acknowledge that there’s room for a 10X or more then 10X gap in productivity. It surprises me how many people claim that there is no difference in productivity among developers. (Why bother practicing and reading blog posts? It won’t make you better!)
I’m more interested in exactly what it takes to turn a median (1X by definition) developer into an exceptional developer.
I don’t buy the trail-blazer and N-1 followers argument because I’ve witnessed massive success (by any metric) cleaning up the non-functioning, non-requirements meeting (but potentially marketable!) untested messes that an unskilled ‘trailblazer’ leaves in their (slowly moving) wake. Do you think it’s all context or are there other forces at work?