I've never seen anyone be 10X better than, say, average consistently over a long period of time. But I've been 10X better than apparently average a few times a

I've never seen anyone be 10X better than, say, average consistently over a long period of time. But I've been 10X better than apparently average a few times a few weeks at a time. Uh, part of the reason for a few weeks at a time is that as part of the 10X I got the work done very quickly! My 10X cases were all when the planets happened to line up just right and, to use a baseball analogy, the work to be done was like a slow pitch right in my best strike zone. (1) I was seeing a girl, and she was an intern and working for someone who wanted her to work on some old Fortran code. She couldn't figure out why the code was giving totally garbage results. I glanced at her code and saw where she passed the constant 2 as an argument to a subroutine. So, sure, that can be dangerous. I glanced at the subroutine and, yup, it was changing the corresponding parameter so for the rest of the execution of the program the constant 2 had the value given it by the subroutine and was not 2. She worked a day or two on this, and I solved it in about 180 seconds. Why? I'd seen that in Fortran usage before. (2) I was in a software house working as a programmer and fast Fourier transform expert at a US Navy lab. Some engineers at the lab needed an extensive software project done and set up a competitive bidding situation. For one of the engineers, part of the work to be done was some power spectral estimation of ocean wave noise, that is, quite low frequency. Due to some earlier experience at GE, I knew the basics of the Wiener-Khinchin theorem and also got the Blackman-Tukey book on the statistics of power spectral estimation, typed in some code to generate synthetic ocean waves and accumulate and display the power spectra estimates, then called the engineer and we met and reviewed my work. The main point I had for him was that for the very low frequencies he was interested in, he could need dozens of hours of data and with much less data than that would be looking at small sample statistical fluctuations instead of good estimates of the actual power spectrum. Likely I got these results in 10X. Our software house got "sole source" on the work. (3) I was at Georgetown University consulting in applied math and teaching computer science courses when a friend from college called. He flew up to Georgetown with several other people, and we met. Everything was inconclusive -- no one had any definite, good ideas. The problem was scheduling the fleet for FedEx. That was important because some Members of the BoD questioned if the fleet could be scheduled at all; crucial funding was being held up; the company was at risk of going out of business. Finally in the meeting I just blurted out that I would solve the problem. Six weeks later I had the software running which was also the end of the semester for my teaching. I flew to Memphis, ran my program, pleased the BoD, pleased the founder F. Smith ("solved the most important problem facing FedEx"), and saved the company. I suspect that in time, that was a 10X effort. (4) I was working on a Navy project to support both my wife and myself through our Ph.D. degrees when the Navy wanted an evaluation of the survivability of the US SSBN (ballistic missile submarines) under a scenario of global nuclear war limited to sea. They wanted the results in two weeks. That timing just fit -- my wife had a vacation planned for us at Shenandoah just after the two weeks. From a WWII paper by Koopman's, I saw a continuous time, discrete state space Markov process subordinated to a Poisson progress and wrote and ran the software, Monte Carlo. The software was fast enough that I could generate and average 1000 sample paths quickly. During a peer-review, an objection was that there was no way I could "fathom the enormous state space". In response I mentioned that after, say, five days, the number of SSBNs left was a random variable; it was bounded and thus had an expectation; the expectation was finite; the random variable had a finite variance; the law of large numbers applied; and averaging 500 sample paths would give quite accurate estimates. I explained that intuitively Monte Carlo put the effort where the action was, and the reviewer, a well-known probabilist, responded "That's a good way to look at it." My work passed the review. A second person, actually two people, had been assigned as an independent effort. At the end of the two weeks, they made no progress at all. The Navy got their results on time, and my wife got her vacation on time. (5) When I arrived at Georgetown, a prof had just completed a statistics package. One of the routines was polynomial fitting, but on testing the numerical accuracy was poor. The prof had programmed normal equations, and on that problem the normal equations are close to the notoriously ill-conditioned Hilbert matrix. I wrote a replacement routine using some orthogonal polynomial methods and got accurate results. (6) I was in an AI (artificial intelligence), expert systems, project at IBM's Watson lab, and we were developing IBM's language YES/L1. We wanted rule subroutines. Our code design was to (A) return one call at a time from the whole stack of dynamic descendancy, (B) execute the RETE algorithm run-time code, (C) rebuild the stack of dynamic descendancy. I saw serious semantic problems with this approach and the coding was to take some weeks. I got some dinner, wrote and ran some illustrative code, in effect to call back into the stack of dynamic descendancy, at dawn wrote email to our team, got some sleep, returned at noon, and our programmer on the work was done later that day. I got an IBM award for the work. There were a few more 10X examples. I always put my pants on one leg at a time and have never walked on water in warm weather. None of these examples involved high stress. Instead the keys were context, e.g., I had everything I needed and background so that the work was in my strike zone. I don't achieve 10X all the time. Heck, I don't achieve 1X all the time; e.g., in my startup recently I've been interrupted and slowed by some unpredictable exogenous obstacles. But now it appears that I've circumvented the obstacles and am making good progress again. Back to it. Net: I don't think anyone can achieve 10X consistently over long time intervals. And I don't believe that very many people can just decide "next week I'll do 10X work." and be able actually to do it. Instead I suspect that most people can have 10X periods given the luck of the right circumstances.