What if you had to design a logo that lasts 10,000 years?

Governments have actually commissioned this, to warn future people away from buried nuclear waste. The maths of erosion, language change and redundancy shows why no single logo can do it, and what can.

Aerial view of the Waste Isolation Pilot Plant surface buildings in the New Mexico desert
Photo: Leaflet, Public domain, via Wikimedia Commons

If you had to design a logo that lasts 10,000 years, you would be doing a real government job: the United States had to plan warning markers for the Waste Isolation Pilot Plant, a nuclear waste repository in New Mexico, to work for 10,000 years. The experts concluded that no single symbol can carry meaning that long, because languages, conventions and even the meaning of a skull change. What can work is a system: many redundant markers, carved deep into low-value stone, using pictures that resemble what they warn about, with messages at several levels of detail.

Key takeaways
  • 7 principles came out of Sandia's 1993 expert panel, from "the site must be marked" to "use materials with little recycle value".
  • 11% to 22% of a language's core words would survive 10,000 years at standard glottochronology rates. Your caption becomes a foreign text.
  • 120 mm of rock surface erodes in 10,000 years at the global mean outcrop rate of 12 m per million years. A carved logo must be deeper than that.
  • 1,650 people in 11 countries were tested in 2007 before the IAEA added a skull-and-running-figure sign, because the trefoil "carried no intrinsic meaning".
  • About 1,400 years: how long Egyptian hieroglyphs went unread after the last inscription in 394 CE.

Brand designers talk about "timeless" logos and mean twenty years. The oldest marks still read today are a handful of cave paintings and scripts that needed a Rosetta Stone to decode. So let me treat the question as an engineering brief, with a real client, a real deadline of 12026 CE, and the physics, linguistics and semiotics that set the limits.

Why would anyone need a logo that lasts 10,000 years?

Because some waste stays dangerous that long. The Waste Isolation Pilot Plant (WIPP) sits about 660 m underground in salt beds near Carlsbad, New Mexico, and has taken defence transuranic waste since 26 March 1999. It is licensed for 10,000 years, and its closure plan includes permanent markers to stop people in the far future from drilling into it.

Finland faces the same brief at a larger scale. In October 2026 its regulator granted Posiva the world's first operating licence for a spent-fuel repository, Onkalo, about 450 m down in bedrock at Olkiluoto, for up to 6,500 tonnes of uranium in roughly 3,300 canisters.

The story

On 13 September 1987, scavengers in Goiânia, Brazil, took an abandoned radiotherapy machine from a closed clinic. There was no danger sign at the site. Inside was a caesium-137 source of about 50.9 terabecquerels. They broke it open, sold parts to a scrapyard, and the owner carried the powder home, where family and neighbours handled it. Four people died, including a six-year-old girl, and 249 were found contaminated among about 112,000 who were checked. Twenty years later the IAEA and ISO finished a five-year programme to add a new warning sign for dangerous sources, because the familiar trefoil, as World Nuclear News put it, carried no intrinsic meaning. Gallup tested candidates on 1,650 people in 11 countries before choosing a skull, a running figure and radiating waves.

Wikipedia: Goiânia accident and World Nuclear News: New symbol ahead (2007)
A yellow and black radiation warning sign with the trefoil symbol on a fence at the Maxey Flat low-level radioactive waste site in Kentucky
The trefoil on a fence at Maxey Flat, Kentucky. To anyone who has not been taught it, the shape means nothing, which is exactly the problem.Photo: RRJackson, CC BY-SA 3.0, via Wikimedia Commons

Why can't a normal logo last that long?

Because almost every logo is what the philosopher Charles Sanders Peirce called a symbol: a sign tied to its meaning only by convention. Conventions are kept alive by people who learn them, and 10,000 years is about 400 generations of teaching.

Peirce sorted signs into three kinds, and the split decides what survives:

  • Icon: resembles its object, like a portrait or a drawing of a person falling.
  • Index: physically connected to its object, like smoke for fire or a weathercock for wind.
  • Symbol: linked by agreement only, like a word, a brand mark or the radiation trefoil.

Icons decay slowest because anyone with eyes can decode a resemblance. Symbols decay fastest. The Nike swoosh, the Apple apple and the trefoil all depend on a living culture to explain them.

A logo is a promise kept by people. Over 10,000 years, the people are the part that changes.From this post

Meanings can flip in a generation

The clearest warning comes from an Indian symbol. The swastika, from Sanskrit svastika, "well-being", has been used for at least 5,000 years and is still sacred in Hinduism, Buddhism and Jainism. In Europe it was a common good-luck sign in the early 1900s. The Nazi Party adopted it in 1920, and within about 25 years its meaning in the West had turned to terror. Five thousand years of meaning reversed in one generation, without the shape changing at all.

How fast does the language around a logo decay?

Fast enough that any words on your marker will be foreign. Linguists model the loss of core vocabulary with glottochronology: Morris Swadesh estimated that about 14 of 100 basic words are replaced per millennium, and Robert Lees's 1953 study of 13 language pairs gave a retention rate of 0.805 per millennium on a 200-word list.

  1. Model. If a fraction \( r \) of core words survives each millennium, after \( t \) millennia the surviving fraction is \( c = r^{t} \).
  2. Swadesh rate. \( c = 0.86^{10} \approx 0.22 \): about 22% of core words remain.
  3. Lees rate. \( c = 0.805^{10} \approx 0.11 \): about 11% remain.
  4. Sensitivity. Icelandic changed slowly, about 4% per millennium: \( 0.96^{10} \approx 0.66 \). Norwegian Riksmål changed fast, about 20%: \( 0.80^{10} \approx 0.11 \).

Calculated in this post from the per-millennium retention rates (0.86 Swadesh, 0.805 Lees; Wikipedia: Glottochronology) raised to the power of 10 millennia.

\[ c = r^{\,t/1000\ \text{yr}} \qquad\Rightarrow\qquad c_{10\,000} = 0.805^{10} \approx 0.114 \]

Even in the kindest case, a reader in 12026 would recognise two-thirds of the basic words of a language like Icelandic, and grammar and spelling would have moved too. Old English, about a thousand years old, already needs a translator. The model is crude: critics such as Sergei Starostin argue rates are not constant and loanwords distort the count. But every version of it says the caption will not be read directly.

The Rosetta Stone in the British Museum, a dark granodiorite slab carved with hieroglyphic, demotic and Greek text
The Rosetta Stone carries one decree in three scripts. Hieroglyphs went unread from the last inscription in 394 CE until Champollion's work in the 1820s. Redundancy across languages is what brought them back.Photo: Hans Hillewaert, CC BY-SA 4.0, via Wikimedia Commons

The scripts that died, and the one that came back

Egyptian hieroglyphs were written from about 3250 BCE until the last known inscription at Philae in 394 CE, a run of over 3,600 years, and then nobody could read them for about 1,400 years. They were recovered only because the Rosetta Stone, found in 1799, repeated one text in hieroglyphic, demotic and Greek. That is the single most useful lesson for a 10,000-year logo: say the same thing several ways.

How deep would you have to carve it?

Deeper than you think. Rock surfaces are not permanent; they wear away at measurable rates, and cosmogenic beryllium-10 dating lets geologists measure those rates across thousands of years.

Understanding Earth's eroding surface with ¹⁰BeEric W. Portenga and Paul R. Bierman · GSA Today, 21(8): 4–10 · 2011

The authors compiled beryllium-10 erosion measurements from 450 bedrock outcrops and 1,149 drainage basins worldwide. Outcrops erode at a mean of 12 ± 1.3 m per million years (median 5.4). Igneous rock such as granite averages 8.7 m/Myr, sedimentary rock 20, and temperate climates 25, while polar outcrops are slowest at 3.9. Basins erode far faster, at a mean of 218 m/Myr, because they include soils and river cutting. The limit: these rates average over millennia and do not capture a single storm, frost-shattering or a human with a chisel.

Read the paper (free)

Worked example: the minimum carving depth

Erosion depth is rate times time, \( D = E \cdot t \). One metre per million years is one millimetre per thousand years, which keeps the units easy.

\[ D = E\,t = 12\ \frac{\text{m}}{\text{Myr}} \times 0.01\ \text{Myr} = 0.12\ \text{m} = 120\ \text{mm} \]
  • Granite (igneous mean): 8.7 m/Myr gives 87 mm in 10,000 years.
  • Temperate climate outcrops: 25 m/Myr gives 250 mm.
  • Polar outcrops: 3.9 m/Myr gives only 39 mm.

With a safety factor of two, which is my own assumption for uneven weathering, the lines of a 10,000-year logo on granite in a temperate climate need to be cut about half a metre deep. That is not engraving; it is sculpture. It explains why the WIPP marker designs talk in terms of 25-foot granite pillars and earth berms rather than plaques.

Rock surface lost in 10,000 years, by setting
See the data as a table
Depth eroded ( mm)
Polar climate outcrops39
Igneous rock (granite)87
Metamorphic rock110
Global mean outcrop120
Sedimentary rock200
Temperate climate outcrops250

Source: Portenga and Bierman, GSA Today 2011 (outcrop 10Be erosion rates × 10,000 years) · Mean rates; individual sites vary widely. Multiplication by 10,000 years is mine.

A replica of a painted horse from the Lascaux cave in red and black pigment on rock
A replica of a Lascaux horse. These are icons in Peirce's sense: anyone can still see a horse after about 17,000 years, though nobody knows what it meant to its painters.Photo: HTO, Public domain, via Wikimedia Commons

What did the nuclear waste experts actually recommend?

A layered system, not a logo. In the early 1990s Sandia National Laboratories convened a panel of outside experts to judge how WIPP should be marked. Their report set out principles that any long-lived identity could borrow.

Expert judgment on markers to deter inadvertent human intrusion into the Waste Isolation Pilot PlantK. M. Trauth, S. C. Hora and R. V. Guzowski · Sandia National Laboratories, SAND92-1382 · 1993

The panel produced seven principles: the site must be marked; messages must be truthful and informative; a marker should have multiple components; it should communicate by multiple means, such as language, pictographs and scientific diagrams; messages should have several levels of complexity; materials should have little recycle value; and the international community should keep records of where repositories are and what they hold. It judged that a marker's power to deter declines with time and depends on the intruder's purpose and technology, and that a passive system readable for 10,000 years needed further study. The limit: this is structured expert opinion, not a tested design.

Read the report (free)
Level I: something man-made is here(berms, huge pillars, the shape of the land)Level II: man-made, and dangerous(pictographs, faces of fear)Level III: what, where, when(diagrams, short texts)Level IV: full records(archives, many languages)More people can read the bottom levels; fewer, the top
Message levels, as in the WIPP marker studies. The base must work with no shared language at all; each higher level adds detail for readers who can decode more.

The plan on the ground

The current WIPP plan describes an outer ring of 32 granite pillars about 25 feet tall around a four-mile square, an earthen wall, 16 more pillars, and a roofless granite information room over the waste. Text would be in the six official UN languages plus Navajo, the local language. The final plan is due around 2028, with messages to be finalised by 2033. One proposed text, drafted in the 1993 report, begins: "This place is not a place of honor."

Can redundancy beat time?

Yes, and the maths shows why the experts asked for many markers rather than one great one. If each marker independently survives a millennium with probability \( p \), it survives ten millennia with probability \( p^{10} \), and the chance that at least one of \( n \) markers survives is:

\[ P_{\ge 1} = 1 - \left(1 - p^{10}\right)^{n} \]
  1. Good stone, p = 0.9. One marker: \( 0.9^{10} = 0.35 \). With 48 pillars: \( 1 - 0.65^{48} \approx 0.999999999 \).
  2. Fair stone, p = 0.7. One marker: \( 0.7^{10} = 0.028 \). With 48 pillars: \( 1 - 0.972^{48} \approx 0.75 \).
  3. Poor stone or reuse, p = 0.5. One marker: 0.001. With 48 pillars: about 0.05.

The answer swings from near-certain to near-hopeless on one input, which is why principle six matters: if the granite is worth stealing for buildings, \( p \) collapses and so does everything else. The model also assumes markers fail independently; a single event, such as a later society quarrying the whole site, breaks that assumption. The probabilities here are my illustrative assumptions, not measured values.

Use this to check what you just read: change the timescale, the per-millennium survival and the number of markers, and see what reaches the future.

Will your 10,000-year mark survive?

So what would a 10,000-year logo look like?

Less like a logo and more like a landscape. Pulling the evidence together, here is the brief I would write, and what each line borrows from.

Design choiceWhyEvidence
Icons, not symbols: people fleeing, a body fallingResemblance needs no teachingPeirce's sign types; ISO 21482 skull and running figure
Same message in pictures, diagrams and 7 languagesOne channel will surviveRosetta Stone; WIPP principle 4
Cut at least 250–500 mm deep in graniteOutlasts 87–250 mm of erosionPortenga and Bierman 2011
Dozens of markers, spread outRedundancy multiplies survival oddsWIPP's 48 pillars; the calculation above
Worthless materialsNobody quarries themWIPP principle 6
Readable in relief, no colourPigment fades on exposed stoneLascaux survived only sealed
Surface buildings of the Onkalo spent nuclear fuel repository at Olkiluoto, Finland, among trees
Onkalo at Olkiluoto, Finland, licensed in October 2026. Finland's approach leans on hiding the site deep in bedrock rather than on marking it loudly.Photo: kallerna, CC BY-SA 4.0, via Wikimedia Commons

Mark it, or hide it?

There is a counter-argument. A huge, strange monument may attract the curious, the way pyramids attracted tomb robbers. Finland has leaned towards depth and records, and the Sandia panel itself accepted that markers deter less over time. My reading of the evidence is that the answer is both: deep isolation as the real protection, and markers as the second line for anyone who digs anyway.

Icon: a person fleeingSymbol: a trefoil-like markMeaning must be taught
An icon resembles what it means; a symbol only stands for it by agreement. Over millennia, agreements are forgotten first.
Verdict: possible with changes, as a system rather than a logo

A single conventional mark cannot be guaranteed to mean anything in 10,000 years: at standard glottochronology rates only 11–22% of core words survive, and the swastika shows a symbol's meaning can flip within 25 years. A marker system can carry "man-made, dangerous, do not dig" that long if it uses icons rather than symbols, repeats one message across pictures, diagrams and seven languages, cuts features well beyond the 87–250 mm that granite loses in 10,000 years, and spreads dozens of worthless-stone markers so that, at 80% survival per millennium, at least one of 48 almost certainly remains.

What does this teach everyday logo design?

More than most brand books. A 10,000-year brief is just the normal brief with the weak points magnified.

  • Lean on resemblance. Marks that look like something age more slowly than pure abstractions, which is one reason so many logos look dated once their style era passes.
  • Design for one colour and for relief. A mark that survives etching, embossing and a fax will survive a rebrand. This is also why a vector master matters, as covered in vector vs raster for logos.
  • Write the rules down at several levels. A one-line purpose, a page of usage and a full manual are Level I to IV for a brand; see my brand guidelines guide.
  • Expect meaning to drift. India's own logo redesigns show how quickly context moves even over decades.

The full working process for a real-world mark is in the logo design process guide, and the Kala world page gathers the rest of the design writing on this site.

What if next?

Questions people ask

What does "This place is not a place of honor" mean?

It is part of a draft message from the 1993 Sandia report on WIPP markers. It tells future visitors that the site is not a tomb or treasure but a warning, written to be translated into many languages.

Why not just use the radiation trefoil?

Because it is a learned symbol with no built-in meaning. In 2007 the IAEA and ISO added a skull, running figure and radiating waves sign (ISO 21482) after testing it on 1,650 people in 11 countries.

What is the atomic priesthood?

The semiotician Thomas Sebeok proposed, in a 1984 report for the US Human Interference Task Force, that a self-renewing group keep the warning alive through ritual and myth, much as religions carry messages for millennia. Critics worried about its politics and secrecy.

Which logo design lasts the longest?

Simple marks that resemble something, work in one colour, and can be carved or embossed. Abstract marks survive only as long as people keep teaching what they mean.

Sources

  1. Trauth, Hora and Guzowski, Sandia: Expert judgment on markers to deter inadvertent human intrusion into WIPP (SAND92-1382) (1993) (seven principles)
  2. Portenga and Bierman, GSA Today: Understanding Earth's eroding surface with 10Be (2011) (erosion rates)
  3. World Nuclear News: New symbol ahead (2007) (ISO 21482 testing, 1,650 people, 11 countries)
  4. Wikipedia: Goiânia accident (1987 dates, activity, casualties)
  5. Stanford Encyclopedia of Philosophy: Peirce's theory of signs (icon, index, symbol)
  6. Wikipedia: Glottochronology (Swadesh and Lees retention rates, criticisms)
  7. Wikipedia: Egyptian hieroglyphs (394 CE inscription, decipherment)
  8. US Holocaust Memorial Museum: History of the swastika (5,000 years, 1920 adoption)
  9. Wikipedia: Waste Isolation Pilot Plant (depth, opening, marker plan, 2028 and 2033 dates)
  10. Wikipedia: Long-term nuclear waste warning messages (message levels, atomic priesthood, ray cats)
  11. Wikipedia: Lascaux (age and discovery)
  12. World Nuclear News: Posiva given go-ahead to operate Finnish used fuel repository (2026) (Onkalo depth and capacity)
nuclear semioticsWIPP markersOnkaloradiation symbolISO 21482Peirce icon index symbolglottochronologylogo design
Sapta

Sapta (Saptarshi Sarkar) is a mechanical engineer turned marketer in Bengaluru and a solo touring rider. He writes the seven Saptaverse blogs. Work with him or read more of Artblog.

Keep reading in Artblog

From the other blogs