The Extra Second That Computers Would Rather Forget

The strangest second on the clock is named 23:59:60. It does not appear in normal life. Most clocks never show it. Many people will never notice whether it has happened. And yet for half a century, this extra second has been one of the small agreements keeping official time tied to the actual rotation of the Earth. A leap second is not a poetic flourish. It is an engineering compromise. Coordinated Universal Time, or UTC, is based on atomic clocks, whose seconds are far steadier than the Earth itself. The planet, rude old mechanism that it is, does not rotate at a perfectly constant rate. It is affected by tides, the atmosphere, the oceans, the motion of the core, and other restless parts of the physical world. Astronomical time, known as UT1, follows that rotation. UTC follows atomic time but has historically been nudged by leap seconds so it does not drift more than about 0.9 seconds away from UT1. That is the elegant version. The inelegant version is that every so often the world's official timekeeping system has asked digital infrastructure to tolerate a minute with 61 seconds in it. The practice began in 1972, after a ten-second adjustment was made to UTC, and the International Earth Rotation and Reference Systems Service has since announced leap seconds when Earth's rotation and UTC needed to be pulled closer together. NIST explains the basic ritual plainly: positive leap seconds have been inserted at the end of a UTC day, usually in June or December, using the sequence 23:59:59, then 23:59:60, then 00:00:00. There is also a provision for a negative leap second, which would remove a second, though one has never been used. That last detail is where the story gets twitchy. All leap seconds so far have been positive. The software world has had enough trouble with those. A negative leap second would be a new species of clock event: instead of briefly repeating or stretching time, the clock would skip ahead. In 2022, the General Conference on Weights and Measures, the international body behind the metric system, formally noted that recent observations of Earth's rotation indicated the possible need for the first negative leap second, whose insertion had never been foreseen or tested. That is metrology language for: we should perhaps not learn this live. So the leap second is now on borrowed time. At its 27th meeting in 2022, the General Conference decided that the maximum allowed difference between UTC and UT1 will be increased in or before 2035. In practical terms, that means leap seconds are expected to stop being inserted for at least a long stretch. The body asked for a plan that would keep UTC continuous for at least a century, with the details to be proposed for agreement at the 28th meeting in 2026. This is not because anyone discovered that the Earth has become punctual. It is because global systems have become less tolerant of tiny discontinuities. The official resolution names the problem directly: leap seconds create discontinuities that risk serious malfunctions in critical digital infrastructure, including satellite navigation systems, telecommunications, and energy transmission. It also points to another, quieter problem. Different operators have invented different ways to handle leap seconds, and those uncoordinated methods threaten the shared meaning of UTC. The hidden comedy of modern time is that many computers do not simply experience a leap second. They negotiate it. Some systems step the clock. Some smear the leap second across a longer period. Some need careful configuration to avoid mixing smeared and unsmeared time sources. Google says it has used a leap smear since 2008, gradually distributing the extra second across the hours before and after the event instead of making servers swallow an abrupt clock step. Its public description recommends a 24-hour linear smear from noon to noon UTC and notes that smeared time temporarily differs from UTC before realigning. That is a practical solution, but it is also a small philosophical scandal. For a day, two systems can both be doing something defensible and yet disagree slightly about what time it is. One is honoring UTC's leap second directly. Another is smoothing the bump so distributed computing remains calm. The point of clocks, as Google dryly observes in its own explanation, is to read the same time in different places. Leap smearing helps machines, but it also reveals that shared time is a service, not a substance. The costs are not merely theoretical. When the leap second at the end of 2016 arrived, Cloudflare later described how part of its DNS infrastructure ran into trouble because code used a clock source that could move backward. A negative time difference leaked into server selection logic, causing errors until engineers deployed a fix. The postmortem is wonderfully specific in the way infrastructure failures often are: not a civilization-ending drama, just a small assumption about time meeting a rare calendar event at midnight UTC. That is why leap seconds make such good Side Column material. They are tiny enough to seem ridiculous and fundamental enough to touch everything. Your phone, bank transaction, map app, power grid monitor, server log, airline system, and telescope schedule all participate in some version of the same bargain: time must be stable enough for machines, but recognizable enough for humans living under a rotating sky. For astronomers and some scientific users, the connection between civil time and Earth rotation is not decorative. If leap seconds go away, UTC will slowly drift farther from UT1. The change will be tiny for everyday purposes. One second of drift is not a lunch plan ruined by the heavens. But over long periods the difference matters to fields that care where the Earth is pointing. NIST notes that if UTC were no longer corrected, the smaller community needing UT1 could use published corrections instead. That is the likely cultural trade: stop forcing every system to absorb occasional clock discontinuities, and ask specialized users to apply the correction they actually need. There is something revealing in that redistribution of burden. Leap seconds once made the sky's irregularity everyone's problem, whether everyone knew it or not. Ending them would make continuous atomic time the default and put astronomical correction where it is most needed. That is not anti-nature. It is an admission that modern civil time is already an interface layered over physics, law, engineering, and habit. The next few years will decide the details, not the emotional truth. As of the U.S. Naval Observatory's current leap-second announcement, no leap second will be introduced at the end of December 2026, and the last positive leap second was inserted at the end of December 2016. We may already have lived through the final ordinary leap second, though that is not guaranteed until the international process finishes its work. If that was the last one, it left very quietly: a one-second pause tucked into New Year's Eve for UTC, noticed mainly by people paid to keep the world's clocks from arguing. That seems about right. The leap second was always a tiny act of humility before a messy planet. Its likely retirement is a different kind of humility: admitting that the systems built on top of time now need the clock to stop flinching.