The Hidden Science of Everyday Timekeeping
Your body, your government, and your devices each keep a different “now.” As Earth’s rotation shifts, the gap is forcing a global policy reset by 2035.
Key Points
- 1Recognize three competing “nows”: biological rhythms, civil UTC-based law, and machine time engineered for resilience, not philosophical purity.
- 2Understand why UTC, TAI, and UT1 diverge: atomic seconds are uniform, Earth’s rotation isn’t—making leap seconds a fragile patch.
- 3Track the coming policy shift: CGPM Resolution 4 targets changes by 2035, reducing leap-second disruption while letting civil time drift.
A few times a year, the modern world rehearses a small miracle: billions of devices, across every continent, agree on what time it is. Then, without warning, a phone is “wrong” by a second. A server log runs backward. A trading system records events out of order. A power grid monitor starts arguing with itself.
The surprise is not that clocks disagree. The surprise is that we ever believed “now” was a single, stable thing.
Time, as you live it, is layered. Your body keeps one kind of time. Your government declares another. Your phone and the internet negotiate a third—one that is less a fact than an estimate, computed under imperfect conditions and defended against failure and fraud.
And as the planet’s rotation shifts—nudged by processes as profound as the dynamics of Earth’s core and as contemporary as polar ice melt—the layers are straining. In 2022, international timekeepers quietly set the world on a path to change how civil time stays aligned with the spinning Earth, aiming for a new approach in, or before, 2035. The decision, formalized by the General Conference on Weights and Measures (CGPM), is often paraphrased as “ending leap seconds,” though the reality is subtler—and more revealing. It’s governance catching up with physics and the internet.
The most consequential clock change of the next decade isn’t a gadget upgrade. It’s a policy decision about how much the world should care that Earth doesn’t spin on schedule.
— — TheMurrow
Three “nows” we live by—human, civil, and machine
The second “now” is collective: civil time, the legal time your society uses for contracts, schedules, courts, and trains. Nearly everywhere, civil time is based on Coordinated Universal Time (UTC) plus a time-zone offset and daylight-saving rules. UTC itself is a compromise: it runs at the uniform rate of atomic clocks, yet it has historically been nudged to stay close to Earth’s rotation.
The third “now” is operational: machine time, the time your phone, your bank, and your cloud provider use to coordinate complex distributed systems. Here, “now” is computed from time sources (like NTP servers and GNSS/GPS signals), local oscillators, and algorithms designed to handle delay, jitter, outages, and even attacks. Machine time is engineered to be good enough to keep systems coherent, not philosophically pure.
Why disagreements happen—even when nobody is “wrong”
Four reasons clocks diverge
- ✓Physics: simultaneity is not absolute, and signals take time to travel.
- ✓Geophysics: Earth’s rotation is irregular; “astronomical” time drifts.
- ✓Engineering: every distribution path adds error; systems trade accuracy for resilience and security.
- ✓Governance: international standards must balance scientific fidelity with global operational reality.
Readers often feel time disagreements as personal annoyance: a phone that’s a second off, a microwave that drifts. The deeper story is systemic. Our most important clocks are not isolated instruments—they’re agreements under pressure.
The source of truth: TAI, UTC, and UT1 (and why they diverge)
TAI (International Atomic Time) is a continuous timescale built from an ensemble of atomic clocks. It counts uniform SI seconds without interruption. No leap seconds. No pauses. No skips.
UTC (Coordinated Universal Time) is the world’s civil reference. It follows the atomic rate—so its seconds match SI seconds—but it has historically been adjusted to track Earth’s rotation closely. The BIPM (International Bureau of Weights and Measures) describes UTC as differing from TAI by an integer number of seconds. That integer reflects accumulated adjustments. (BIPM)
UT1 is the Earth-rotation side of the ledger: a measure tied to Earth’s actual rotation angle, used in astronomy and Earth orientation.
Here’s the tension: atomic time is smooth; Earth’s rotation is not. Over long enough periods, UT1 and UTC want to drift apart, unless you intervene.
Leap seconds: a practical patch with a growing bill
That sounds gentle: a single second, occasionally. In practice, it’s brutal for systems that assume time is monotonic and uniform. Many programs treat “time since epoch” as always increasing by one second per second, forever. Leap seconds violate that assumption in ways that surface in logs, ordering guarantees, authentication windows, and real-time coordination.
Atomic clocks keep time. Leap seconds keep politics—and software—on speaking terms with the planet.
— — TheMurrow
Leap seconds also carry a technical asymmetry: most software has been written assuming leaps are positive—an extra second added—because that’s all we’ve ever used. The possibility of a negative leap second forces developers to confront an edge case they trained themselves to ignore.
Why leap seconds became a political problem, not just a technical one
Digital infrastructure is the backdrop to the dispute. Global systems need timestamps that behave predictably across data centers and networks. When a leap second arrives, engineers must choose between conflicting ideals:
- represent UTC “correctly,” including its occasional discontinuities, or
- enforce monotonic time to preserve ordering and avoid failures.
Many operators handle the dilemma with a compromise—spreading the leap second across a window by “smearing” time—so clocks never show a repeated second. That makes local systems happier but means different places can temporarily disagree about UTC. Civil time remains nominally unified while machine time becomes, in practice, negotiated.
The governance angle is where the story sharpens. In November 2022, the CGPM adopted Resolution 4, deciding that the maximum permitted difference (UT1 − UTC) will be increased in, or before, 2035. The resolution also instructs that a plan be prepared and brought to the 28th CGPM (2026). (BIPM)
Policy watchers described this as a path toward ending leap seconds “in their current form”—not necessarily abolishing adjustments forever, but deferring and restructuring them so civil time can run uninterrupted for longer stretches.
Two legitimate perspectives, one hard compromise
Defenders of tight coupling to Earth’s rotation argue from meaning and continuity. Civil time historically tracks the Sun’s rhythm; untethering the clock from the sky feels like a philosophical and cultural break. UT1 matters for astronomy and satellite tracking, too, though those domains already use specialized timekeeping.
The CGPM resolution reads like an admission that the status quo is too expensive to maintain. Not in money, necessarily—in fragility.
The leap-second debate is what happens when the oldest clock we have—the rotating Earth—meets the newest kind: the planet-scale computer.
— — TheMurrow
The Earth won’t hold still: rotation, uncertainty, and the “negative leap second” fear
A 2024 paper in Nature by Duncan Carr Agnew (UC San Diego/Scripps) argues that if recent trends continued, the current UTC definition could require a negative leap second by ~2029—a type never used before. (Nature)
A negative leap second would mean skipping a second to bring UTC back toward UT1. It sounds trivial until you translate it into system behavior: many implementations assume leap seconds only ever add time, never remove it. A skipped second can be nastier than a repeated one, particularly for:
- event ordering and log consistency,
- time-based authentication and expiry logic,
- systems that schedule actions at exact timestamps,
- real-time monitoring that assumes each second exists.
Nature’s news coverage emphasized that negative leap seconds are uniquely scary because our software ecosystem has trained itself on one kind of discontinuity. (Nature)
Climate, ice, and the clock: what the paper claims—and what it doesn’t
That’s not a headline about “climate change breaks time.” It’s a window into how interconnected systems are: mass moving across a planet changes rotation; rotation influences time standards; time standards shape global computing.
It’s also not a guarantee. Coverage and expert commentary stress uncertainty: Earth’s rotation—especially contributions from the core—is hard to forecast precisely. The Washington Post noted meaningful uncertainty in predictions even while emphasizing the increased plausibility of a negative leap second compared to earlier expectations. (Washington Post)
For readers, the most honest takeaway is probabilistic: the risk profile shifted. Engineers and standards bodies are now planning around a scenario that once lived in the footnotes.
Key Insight
How your phone decides what time it is (and why it can “agree” and still be wrong)
Modern devices synthesize time from multiple sources: network time protocols, satellite signals, cellular timing, and local oscillators. Each source arrives with imperfections. Signals have delay. Networks have jitter. Receivers have noise. Attackers can spoof.
So device time becomes an engineered estimate. Good systems do three things well:
1. Choose sources wisely (prioritizing stable, trustworthy signals).
2. Model uncertainty (recognizing that incoming time stamps are not perfect).
3. Avoid dangerous jumps (because sudden corrections can break apps and logs).
What good device timekeeping does
- 1.Choose sources wisely (prioritizing stable, trustworthy signals).
- 2.Model uncertainty (recognizing that incoming time stamps are not perfect).
- 3.Avoid dangerous jumps (because sudden corrections can break apps and logs).
A real-world example: when “correct” time loses to “safe” time
The same tension appears with leap seconds. Some systems “smear” them to keep monotonic time. Others represent them explicitly. Both are defensible. The point is that “right time” is not the only priority; “time that doesn’t break everything” often wins.
For readers, the practical implication is simple: disagreement by a second doesn’t automatically mean malfunction. It can be the signature of a system choosing stability over strict alignment.
Case study: the leap second as an infrastructure stress test
The systems most sensitive to time discontinuities share a few traits:
- they operate at large scale (many machines, many regions),
- they coordinate events that must be ordered consistently,
- they embed assumptions about monotonic clocks,
- they rely on logs for audit, recovery, or forensic analysis.
A leap second—especially an unexpected or poorly handled one—can produce confusing symptoms: duplicated timestamps, “impossible” event sequences, and cascading alerts. Even when nothing catastrophic happens, the operational burden is real: planning, testing, monitoring, and post-event verification.
The negative leap second scenario is worse because it asks infrastructure to handle a discontinuity that most implementations have never exercised. A positive leap second can be represented as a repeated timestamp. A negative leap second demands a missing timestamp—an absence that some systems don’t know how to express.
That’s why the 2022 CGPM decision matters beyond metrology circles. Increasing the allowed UT1 − UTC difference by 2035 aims to reduce the frequency of these stress tests, buying the world longer periods of uninterrupted UTC. (BIPM)
The cost is conceptual drift: civil time would slowly decouple from Earth rotation more than before. The benefit is operational calm.
Infrastructure takeaway
What the 2035 shift could mean for daily life—and for the systems underneath it
The more meaningful changes will be institutional and technical.
Practical implications for readers
- A slower drift between clock time and astronomical time: The sky won’t change, but the definition of civil time’s relationship to Earth’s rotation will loosen.
- More explicit separation of civil vs astronomical time: Fields that need Earth rotation (astronomy, some navigation and space operations) already use UT1-like references. A larger permitted UTC offset formalizes the split.
A governance lesson hiding in a timestamp
Resolution 4’s timeline is itself a statistic worth sitting with:
- 2022: CGPM adopts Resolution 4. (BIPM)
- 2026: plan to be prepared for the 28th CGPM. (BIPM)
- 2035: change to the maximum allowed UT1 − UTC difference in, or before this year. (BIPM)
Those dates reflect how hard it is to change time without breaking the world. Timekeeping isn’t only about precision. It’s about coordination across governments, industries, and the everyday habits of billions of people.
TheMurrow take: “Now” is a negotiation—and that’s not a failure
The real achievement is that our time system works at all, given the constraints. Physics denies us instant synchronization. Geophysics denies us a perfectly stable day. Engineering denies us perfect distribution. Governance denies us unilateral choices.
So “now” becomes a negotiation: between atomic regularity and astronomical meaning, between monotonic machine time and civil expectations, between the planet we inhabit and the networks we’ve built on top of it.
The coming shift—formalized in 2022 and targeted for implementation by 2035—signals a world choosing operational robustness over tight adherence to Earth’s rotation, at least for civil time. The debate will continue, because the trade-off is real. Yet the direction is clear: the internet’s needs have become a forcing function for how humanity defines the present.
If that sounds abstract, try a simpler thought. Your body’s “now” can be jet-lagged. Your government’s “now” can change with legislation. Your phone’s “now” is computed and defended.
None of them is the whole truth. Together, they are how we keep appointments with one another on a moving planet.
Frequently Asked Questions
What’s the difference between UTC and TAI?
TAI (International Atomic Time) is a continuous atomic timescale with no leap seconds. UTC (Coordinated Universal Time) runs at the atomic rate but has historically been adjusted so it stays close to Earth-rotation time. According to the BIPM, UTC differs from TAI by an integer number of seconds, reflecting those adjustments.
What is UT1, and why does it matter?
UT1 tracks Earth’s actual rotation angle—effectively the astronomical “how far has Earth turned” time reference. It matters for applications tied to Earth orientation, including astronomy. UT1 is also the reason leap seconds exist: without adjustments, UT1 − UTC would drift as Earth’s rotation varies.
Why were leap seconds introduced in the first place?
Leap seconds were introduced to keep UT1 − UTC within a tolerance (historically near 0.9 seconds), so civil time stayed aligned with Earth’s rotation. Earth doesn’t rotate uniformly, so the difference accumulates unless UTC is adjusted. The adjustment historically took the form of adding a second occasionally.
What is a negative leap second, and why are people worried about it?
A negative leap second would mean skipping a second to bring UTC closer to UT1. It has never been used. A 2024 Nature paper by Duncan Carr Agnew suggested one might be needed around 2029 if trends continued. The worry is technical: many systems were built assuming leap seconds only ever add time, not remove it.
Is climate change really affecting timekeeping?
The 2024 Nature analysis argues that polar ice melt redistributes mass on Earth, affecting rotation and influencing when a negative leap second might be needed. The paper suggests melting could delay that need by about three years compared with a scenario without that effect. Predictions carry uncertainty because Earth’s rotation is difficult to forecast precisely.
What did the international community decide about leap seconds in 2022?
In November 2022, the CGPM adopted Resolution 4, deciding that the maximum allowed difference (UT1 − UTC) will be increased in, or before, 2035. The resolution also calls for a plan to be prepared for the 28th CGPM (2026). The change is widely described as a move away from leap seconds in their current form.
