(hide navigation)
  • Swedish content
Fund my projects
Don't miss
Page thumbnail
Phasor, high quality video
Log in
Latest comments
RSS feed

Kernighan's lever

Brian Kernighan famously wrote:

Everyone knows that debugging is twice as hard as writing a program in the first place. So if you're as clever as you can be when you write it, how will you ever debug it?

The Elements of Programming Style, 2nd edition, chapter 2

The following version also circulates on the net:

Debugging is twice as hard as writing the code in the first place. Therefore, if you write the code as cleverly as possible, you are, by definition, not smart enough to debug it.

This second quote may or may not be by Kernighan — the questionable use of "by definition" makes me uncertain — but it is useful as a provocative sound bite conveying the same essential idea.

It is tempting to interpret Kernighan's aphorism as a warning: Stay away from clever techniques, it seems to say, because if you write clever code, you will never be able to get it to work. But this interpretation is unfortunate, and rests on the false assumption that cleverness is static.

While it is possible that Kernighan intended us to interpret the message in a specific way, he wisely restricted himself to merely presenting an observation, allowing us to draw our own conclusions from it.

Pay close attention to what is actually being said: Having written code as cleverly as you can, you will suddenly face a problem that you are not clever enough to solve. Certainly, "clever" in this context does not refer to some innate talent, because nobody is born with the ability to write clever code in the first place. The "cleverness" required to write and understand intricate code is an acquired mental skill.

Cat at keyboard

If you are a programmer, you will be familiar with a sense of wonder, gradually transforming into utter stupor, as you stare at some perfectly reasonable code that couldn't possibly fail, and yet somehow it does. And since you are confident that you understand how the code works, having written it yourself, you feel that you must be able to figure out what is going on. Not only the desire to deliver working software on time, but other powerful forces such as pride, stubbornness and curiosity contribute to the motivation that pushes you onwards through the arduous task of tracking down the root cause of the error. Suddenly you see it, and you're blinded by a bright light as all the pieces fall into place. The inexperienced programmer may fall into the trap of self-degradation: "Oh, look at how stupid I was!" But that same sentiment is proof that your programming-related cleverness, or skill, has increased: "Oh, look at how clever I've become!" (Although I wouldn't recommend saying that out loud.)

Skill is the result of practice, that is, of systematically trying to work slightly beyond one's ability. Quite understandably, most of us don't spend that kind of effort unless we have good reason to. Hence, without motivation we do not practise, but simply cruise along at our current level and never improve any further.

The mind is very good at rationalising, and will convince us that our current skills are sufficient, that we are all "good enough"; certainly better than the average programmer anyway. The human brain will do this trick regardless of our actual level of skill. So while we all tend to consider ourselves sufficiently skilled right now, we never regret improving.


You effortlessly wield clever programming techniques today that would've baffled your younger self. (If not, then I'm afraid you stopped evolving as a programmer long ago.) But this improvement is the result of practice, and something must have motivated you to put in all those hours of work. Kernighan's witty remarks provide a clue: In programming, as soon as you work at your current level, you will automatically end up in a situation where you have to work beyond your current level. By means of this very fortunate mechanism, you will leverage several basic human drives (honour, pride, stubbornness, curiosity) into providing the motivation necessary for improvement.

I call this mechanism Kernighan's lever. By putting in a small amount of motivation towards the short-term goal of implementing some functionality, you suddenly end up with a much larger amount of motivation towards a long term investment in your own personal growth as a programmer.

If we deliberately stay away from clever techniques when writing code, in order to avoid the need for skill when debugging, we dodge the lever and miss out on the improvement. We would then need other sources of motivation in order to grow as programmers, and if no such motivation appears, our abilities stagnate (or even deteriorate).

The psychological concept of flow, somewhat simplified, can be used to visualise the process. Flow is when you are "fully immersed in a feeling of energised focus, full involvement, and enjoyment in the process of the activity" (wikipedia), and it only occurs when the challenge that you are tackling matches your current level of skill.

Flow chart

Implement below your ability, and you get to debug in the "flow" area.

Flow chart

Implement at your ability, and the debugging will be frustrating, but you gain skill.

You will find yourself situated at a particular x-coordinate, corresponding to your current level of skill. If writing code is a point on this graph, then (according to Kernighan's assumption) debugging the same code would be a point a fair bit directly above it.

It is certainly possible to deliberately pick a low starting point just to avoid ending up in the frustration area. But this will put you squarely in the boredom area, and boredom is frankly no better than frustration. However, should you pick a starting point in the enjoyable flow area, Kernighan's lever will screech into action and push you sideways through the graph, increasing your skill until it matches the challenge posed by the bugs in your code.

Naturally, the real world is more complex than this, and you will sometimes have compelling reasons to go for the boring option, and artificially reduce your cleverness in order to dumb down the debugging phase. But it may harm your long-term personal development if you go down that road every single time you write a program.

In conclusion, the answer to Kernighan's rhetorical question — "how will you ever debug it?" — is straight-forward: By tackling the problem, thereby gaining valuable experience and becoming more clever in the process. And the second version of the quote can be adorned with a single word at the end: Yet.

Posted Tuesday 11-Dec-2012 21:33

Discuss this page

Disclaimer: I am not responsible for what people (other than myself) write in the forums. Please report any abuse, such as insults, slander, spam and illegal material, and I will take appropriate actions. Don't feed the trolls.

Jag tar inget ansvar för det som skrivs i forumet, förutom mina egna inlägg. Vänligen rapportera alla inlägg som bryter mot reglerna, så ska jag se vad jag kan göra. Som regelbrott räknas till exempel förolämpningar, förtal, spam och olagligt material. Mata inte trålarna.

Sat 22-Dec-2012 00:10
If this is all true, a perfect solution would be to debug the program 10 years after you've wrote it...

However, you will be tempted to rewrite it because after those years the code will be rated down into the boring area, so you'll need another 10 years to wait to debug it.
Linus Åkesson
Sat 22-Dec-2012 14:09
If this is all true, a perfect solution would be to debug the program 10 years after you've wrote it...

No, that's completely contrary to my point. Was I really that unclear?

Work at your current capacity for ten years and your skills will improve only slightly. Spend ten years trying, with great effort, to work beyond your current capacity, and your skills will improve a lot. The tricky part is motivating yourself to spend that effort. This is common knowledge.

My article is about how software developers get this motivation for free by writing clever code.
Max Porshnev
Tue 25-Dec-2012 13:34

lft wrote:

Spend ten years trying, with great effort, to work beyond your current capacity, and your skills will improve a lot.
A famous russian singer <a href="http://en.wikipedia.org/wiki/Vladimir_Vysotsky">Владимир Высоцкий</a> had a song about that: "лучше гор могут быть только горы, на которых ещё не бывал". The song is called "Прощание с горами".
Fri 11-Jan-2013 13:06

mporshnev wrote:

"лучше гор могут быть только горы, на которых ещё не бывал"

Or, loosely translated to English, "Better than a mountain can only be mountains you have not scaled yet".
Sun 27-Jan-2013 01:33
You just put delicate words onto an everyday personnal feeling. And the tricky part about motivation is not an issue IMHO, because there is no biggest reward than feeling improvement of one's skill. This is self-motivating. Having an idea and code it in the boredom area in order to avoid frustration is like treating your own idea with disrespect. I prefer writing code that respects any original idea thus requiring full capability, which will likely lead to bugs and so on, as you said, but also, hopefully, to a point where you honored your idea (and fixed your bugs).
Sun 5-May-2013 19:46
Interesting essay with some valid points, but the context lacks depth: it is one thing to do this because it is a hobby, quite another if one must walk the very delicate line of writing code easy to debug in consideration of others, because one is an entrepreneur, racing against the clock. "When are we going to ship?"

You are probably privy to the Zen of "software is done when it is done!", which means that writing quality code must necessarily avoid cleverness in favor of maintainability: if one is an entrepreneur, one must eventually ship, or the startup will bust.

Rule of clarity: clarity is better than cleverness

"Because maintenance is so important and so expensive, write programs as if the most important communication they do is not to the computer that executes them but to the human beings who will read and maintain the source code in the future (including yourself)."


Rule of representation: fold knowledge into data, so program logic can be stupid and robust

"Even the simplest procedural logic is hard for humans to verify, but quite complex data structures are fairly easy to model and reason about. To see this, compare the expressiveness and explanatory power of a diagram of (say) a fifty-node pointer tree with a flowchart of a fifty-line program. Or, compare an array initializer expressing a conversion table with an equivalent switch statement. The difference in transparency and clarity is dramatic."

Sun 5-May-2013 19:53
I would also like to leave you to think about the following:

anyone can write clever code; anyone can write complex code; anyone can over-complicate...

...but if you are really, really smart, make a complex thing simple! Can you go the other way, can you effectively simplify complex things? That is true smarts, because it is incredibly hard to do, and requires one to be extremely smart, knowledgeable, and experienced. Not very many people can do it, even among the smart.
Sat 8-Jun-2013 04:27
"Everything should be made as simple as possible, but not simpler." - Einstein
Sun 25-Aug-2013 23:07
I do not think it is necessary to deliberately stay away from clever techniques. It depends what you do. Also, debugging isn't exactly twice as difficult; it depends on the program and on other things.
Mon 28-Jul-2014 06:51
It true that debugging do improve the understanding of the myth underlying the technologies, especially when debugging others's code in programming.
Mon 28-Jul-2014 07:01
Very nice. Good of you to not attempt to debunk Kernighan but leverage his quote!
Fri 8-Aug-2014 03:36
While practice is generally useful, some practice is more useful than other practice. We are not trying to maximize practice, we are trying to maximize skill-building per time, subject to the motivational constraints. I believe debugging is often less skill-building than writing code - for instance, an hour-long hunt for an omitted character often causes little or no skill enhancement. Sadly, the laws of physics do not say "you will gain skill in accordance with how much you work".

Also, I find dubious your statement that coding skill is not innate because "because nobody is born with the ability to write clever code in the first place". "Practice improves programmer skill" does not imply "*only* practice improves programmer skill". Try telling a 3-foot-tall man that the ability to play basketball isn't innate, because "nobody is born knowing how to play basketball".
Fri 1-Apr-2016 09:41
[KC] I love this Kernighan quote. As the author of the best language and best programming book ever, he's one of the computing greats who will be remembered forever.

Nice article too Linus. I think it sure takes twice the effort to debug, for the initial effort in analysis and then implementation has to be revisited, before the problem can be seen in context of the original solution.

"That can't possibly be happening, the code just cannot do that!". I love problems that seem so impossible. The more people tell me that, the more obvious I know the solution will be when we find it.

Subtle problems are much more difficult to find :)

From my own (30yrs+) experience in commercial software, Kernighan is right.
Tue 7-Jun-2016 22:26
3 aphorisms (aphorisi?), all related, that i've collected over the long, arduous years of prof s/w, all a testament to BK's thesis:

1. if builders built buildings the way programmers build programs, the first woodpecker that came along would destroy civilization.

2. nelson's law: the better the four-wheel drive vehicle, the farther away from civilization you'll be when it breaks down.

and 3, the corollary: the probability that a bug exists is directly proportional to the programmer's insistence that it does not.

thank you for a great article (and the accompanying apropos pic of archimedes)
Andrea "6502" Griffini
Tue 10-Apr-2018 08:31
Something that programmers should try to do in my opinion is

1) write code at your limits, trying to push them farther
2) but don't do that in *production* code :-)

Code in production should be "trivial"... even if of course what is trivial and what is not depends on the reader.

Moreover the problem of writing the simplest code that can do the job can be interesting too (keeping boredom away), even if of course at a different level than writing the smartest, the fastest or the smallest.

However I don't think there is such an easy answer to the question of code complexity... the only definitive answer I think there is in programming is "it depends", and the simplest code may be too simple to be future proof and not able to accept easily the next feature that will be required, forcing a rewriting.

Nothing bad in rewriting itself of course (au contraire), but may be the economic downside won't make that an acceptable philosophy.
Thu 1-Oct-2020 23:55
Wow this page is used in our university programming course.
Thu 18-Aug-2022 19:41
If this is all true, a perfect solution would be to debug the program 10 years after you've wrote it...

However, you will be tempted to rewrite it because after those years the code will be rated down into the boring area, so you'll need another 10 years to wait to debug it.

Wed 21-Dec-2022 20:41
Mark Fitzsimmons
Engineer, woodworker, philosopher, fan
My response is, both of the original statements are false, because the skills required for programming and debugging are subtly different, and not inherently easier or harder. The debugging task has methods that can be systematic and stepwise (making the task "easier"), and sometimes requiring additional code writing to interrogate what were, in your original inspired creation phase, leaps of intuition or complex design you imagined were simpler than you thought (making the task "harder"). Debugging is in part a process of learning how complex your design intent really is, when translated into a language of a complex machine. They are not really separate things, programming and debugging.
Tue 15-Aug-2023 07:17
This matches with my experience as well. Thanks for sharing.

Sun 11-Feb-2024 22:21
My interpretation of "clever" as Kernighan uses it is based on writings by him and Thompson in the Unix heyday. Thompson wrote the kernel using simple mechanisms that could be optimzied later if required, and only optimized after profiling. So "clever" could be interpreted as overdesigned. Of course the points you make about improvement are valid, and if we interpret "clever" as meaning as good an algorithm and data structure as possible, then yes, by all means, so long as the pattern can be discerned by others later. Like good writing, one must write elegantly and clearly.