My Own, Because I Can

illustration

The salon’s administrator asks the agent to let clients log in to the website with their own password. The agent manages in about five minutes: it writes a login form, works out how to store the passwords, writes the check, and everything works. The administrator is pleased, the agent is pleased, and neither of them asks why this task required writing something of one’s own, when the world has dozens of ready-made, tested, repeatedly hacked and repeatedly patched login systems that do this and do it well.

A home-made login system may very well store passwords slightly wrongly, and the mistake will be invisible and of such a size that a year later, when the salon’s database is stolen along with the databases of a hundred other small websites, the clients’ passwords can be recovered in an evening and tried on their email and banking apps. In 1998 the cryptographer Bruce Schneier formulated a rule that has since borne his name: anyone, from the most clueless amateur to the best cryptographer, can create an algorithm that he himself cannot break. Creating a cipher that nobody else can break even after years of attempts is immeasurably harder. Charles Babbage wrote something similar as early as 1864, so the observation is a century and a half old, and in a century and a half it has not once gone out of date.

The desire to write your own where something ready-made exists is called by programmers the not-invented-here syndrome, or NIH for short. Language models are at the same time finishing this disease off and making it more dangerous than ever.

---

illustration

The syndrome was studied long before AI. In 1982 the researchers Ralph Katz and Thomas Allen published a study of fifty research groups in one large laboratory and looked at how their performance depended on how long people had worked together. At first it rose, up to about a year and a half of working together, then for a while it did not change, and by five years it fell noticeably. The explanation lay in whom the groups talked to: the longer a group existed with the same members, the less often they communicated with each other and with outside sources of information, and the more firmly they believed that they themselves knew everything important in their field. Katz and Allen defined the syndrome in those terms: the tendency of a stable group to believe it has a monopoly on knowledge in its field and to reject ideas from outside, to the detriment of its own work.

The syndrome also has a psychological lining, which in 2012 was measured by three researchers, Michael Norton, Daniel Mochon and Dan Ariely. They had people assemble a box from IKEA, fold an origami figure or build something from a construction kit, and then asked how much they thought the result was worth. People valued their own clumsy creations about as highly as the work of experts and expected others to share that valuation. The effect disappeared if the person did not finish the creation or dismantled it himself afterwards. The researchers called it the IKEA effect: labour leads to love. Anyone who has seen a programmer defend his code against a proposal to replace it with a ready-made library will recognise the effect at first sight.

---

The most expensive story of the urge to write one’s own happened at the end of the nineties to Netscape, a company whose browser had shortly before been the main door to the internet for most people on the planet. The developers decided that the browser’s old code was too tangled and rewrote it from scratch. This took almost three years, during which the company released no major new versions, while its competitors meanwhile did. By the time the new browser was ready, the market belonged to others.