The measured findings in this group fall into three kinds. There are exponential curves that described decades of technological progress, several of which have now bent or stopped. There are mathematical limits that no amount of progress can overcome. And there are regularities in how software is built, which have held up for half a century precisely because they are about people rather than machines.
The curves and their ends
Moore's Law, Gordon Moore's 1965 observation that transistor counts double roughly every two years, is the most famous prediction in technology. Around it sits a family of similar curves: Koomey's Law for computation per unit of energy, Butters's Law and Gilder's Law for communications, Kryder's Law for disk storage, and Swanson's Law for the falling cost of solar panels.
The instructive part is how they end. Dennard Scaling, which let smaller transistors run faster without running hotter, broke down in the mid-2000s, which is why processors stopped getting faster and started getting more cores. Kryder's law slowed sharply after about 2010 as magnetic recording neared its physical limits. Bell's Law of Computer Classes describes what happens instead: roughly every decade a cheaper class of computer appears, from minicomputers to phones, and builds a new industry.
Limits no progress removes
A second set are proofs. Turing's Halting Problem shows that no general procedure can decide whether an arbitrary program will finish. Shannon's Law sets the maximum rate at which information can cross a noisy channel, and the Nyquist–Shannon Sampling Theorem sets how often a signal must be sampled to be reconstructed. The CAP Theorem shows that a distributed database facing a network failure must give up either consistency or availability.
The laws of building software
The third set concerns software projects and has held up for fifty years. Brooks's Law, from The Mythical Man-Month (1975), holds that adding people to a late software project makes it later, because new members need training and communication paths multiply. Lehman's Laws find that software in real use must keep changing and grows more complex unless work is done to simplify it. Wirth's Law observes that software slows down faster than hardware speeds up. Spolsky's Law of Leaky Abstractions and Hyrum's Law find that every simplifying layer eventually exposes what it hides, and that with enough users, every observable behaviour becomes something someone depends on.
Where it applies
The curves are used, and misused, in forecasting. Anyone planning a product or an investment around continued exponential improvement needs to know which curves are still running and which have bent. The theorems set hard boundaries for engineers: the CAP theorem forces an explicit choice in every distributed system, and Shannon's limit tells telecom planners what is physically possible. The software laws apply to every project manager: Brooks's law argues against rescuing a late project by hiring, and Hyrum's law warns that changing any behaviour of a widely used system will break someone.
Where it is argued over
The theorems are settled; the arguments concern the curves. Moore's law has been declared dead many times, and partly survives by redefinition, since the industry treated it as a target to plan around rather than a law of nature. Nielsen's Law was derived from one user's connection history rather than a representative sample. More generally, critics of technological forecasting point out that the curves most often quoted are the ones that have kept going, which makes exponential progress look more reliable than it has been.
Why it is worth knowing
Much talk about technology assumes that progress is smooth and unlimited. This group corrects that picture in three ways: some curves bend, some limits cannot be crossed, and the human side of building software has changed far less than the hardware. Knowing all three is the basis for judging any forecast about what technology will do next.
The entries behind this essay: Evidence in Technology & Software. Every subject has its essays on the essays page.