Socrates distrusted writing

In Plato’s Phaedrus, he recounts the argument that written marks would weaken memory, encouraging people to rely on external records rather than exercise their minds. Readers might acquire the appearance of wisdom without the substance of understanding.

Two and a half millennia later, the technology has changed. The anxiety sounds familiar.

AI will weaken thinking. Students will stop learning. Professionals will lose judgement. Writers will forget how to write. Knowledge will grow shallow because machines can produce answers faster than humans can construct them.

Some of those concerns may prove justified.

Humans have been having versions of this argument for a very long time.

Agriculture may have looked like a terrible idea

Imagine an early farming settlement seen through the eyes of a hunter-gatherer.

The farmers stay in one place. They live in larger communities, often beside domesticated animals, and depend heavily on a narrow range of crops. Infectious disease spreads more easily. A failed harvest threatens everyone at once.

Archaeological evidence suggests that the transition to agriculture coincided, in many populations, with poorer health indicators. Average stature declined in some regions. Dental problems increased. Signs of nutritional stress became more common.

A hunter-gatherer looking across at these communities might reasonably have concluded that the farmers had made their lives worse.

For a long stretch of time, the evidence may even have supported that judgement.

Agriculture transformed humanity for reasons largely invisible at the beginning. Settled food production eventually supported larger populations, specialisation, cities, organised states, written administration, universities and technologies that the first farmers could hardly have imagined.

The costs were real.

The possibilities sat outside the frame of anyone living inside the transition.

That distinction matters.

New systems get measured by old standards

A disruptive technology usually arrives looking inferior on at least some dimension the established system values.

Writing can weaken memorisation because information no longer has to reside entirely inside the human mind.

Printing made copying less exclusive. To scribes trained in manuscript culture, cheap reproduction could look like degradation rather than democratisation.

Early automobiles were noisy, unreliable and dangerous beside a good horse. They broke down on poor roads with little fuel infrastructure.

Personal computers looked trivial beside mainframes. Online shopping looked laughably small beside established retail. Early digital photography offered poorer image quality than film.

Judging an emerging technology by those measures can be perfectly rational at the time.

The error comes when present limitations are projected indefinitely into the future.

Technologies improve. Infrastructure grows around them. Social norms shift. Occupations emerge. Regulation follows. Complementary tools appear.

Eventually the original comparison stops making sense.

Smartphones are rarely assessed as telephones any more because the category itself changed.

Resistance carries information

Technological history is often told as a procession of foolish sceptics.

It is comforting. It is also misleading.

The Luddites illustrate the problem. They are remembered as workers who hated machinery for its own sake. Their objection was considerably more substantial.

Machines were altering wages, working conditions, bargaining power and the value of skilled labour.

The machines worked

That was precisely the problem.

Technological change distributes benefits unevenly. A machine can raise total productivity while destroying the livelihood of the person whose skill it replaces. Society can become richer while particular communities absorb most of the transition cost.

Both things can be true.

Recorded music threatened musicians who lived on live performance. Photography challenged established assumptions about artistic skill. Mechanisation displaced crafts. Spreadsheets removed enormous quantities of manual calculation. ATMs automated one of the central functions of bank tellers.

Many of those professions survived, but their boundaries changed.

New tasks became valuable once machines handled the old ones more cheaply.

Resistance often identifies something real. The useful test is whether it points to a permanent limitation of the technology or a genuine cost of transition.

Those are very different claims.

The skill the machine performs

Calculators produced another familiar argument.

Children who could press buttons for arithmetic might never learn mathematics.

The outcome landed somewhere between prohibition and surrender.

Students still learn arithmetic. Calculators absorb work that would otherwise consume time. Education gradually shifted towards deciding which cognitive skills humans still need to master directly.

Spelling checkers, search engines, navigation systems and spreadsheets followed similar paths.

Few analysts now build complicated financial models by hand merely to prove that they can. A financial analyst still needs enough mathematical understanding to recognise when the spreadsheet is producing nonsense.

GPS reduced the need to memorise routes. It also left many people less capable of navigating without it.

Technology grants capability while sometimes eroding the underlying human skill.

Socrates had a point.

The sharper question is which skills remain valuable enough to preserve deliberately.

AI brings that question to the centre of intellectual work.

Cognition changes the stakes

Earlier machines primarily replaced physical labour, calculation, storage or transmission.

Generative AI drafts, summarises, analyses, translates, codes, illustrates and increasingly reasons across information.

The perceived threat therefore feels more personal.

A machine replacing muscle raises economic questions.

A machine appearing to replace parts of thinking raises questions about human value.

Writers ask whether writing matters when paragraphs appear instantly. Programmers wonder which parts of coding remain distinctively human. Teachers worry about learning stripped of intellectual struggle. Designers face systems capable of producing hundreds of visual alternatives in minutes.

Some jobs will change substantially. Some tasks will disappear. New roles will emerge. Certain skills will lose economic value.

History offers no comfortable guarantee that everyone displaced by technological change eventually benefits from it.

It offers a different lesson.

The first-order effect of a new technology rarely reveals its final social form.

Societies build around technologies

Cars replaced horses and then went much further.

Cities were rebuilt around them. Roads expanded. Petrol stations appeared. Traffic laws developed. Insurance spread. Suburbs grew. Restaurants, tourism and retail reorganised themselves.

The internet replaced letters and then reshaped commerce, publishing, dating, advertising, banking, entertainment, politics and the organisation of work.

Printing made books cheaper and then altered who could distribute ideas, who could challenge authority, and the scale at which scientific knowledge could accumulate.

Agriculture did far more than produce food differently.

Transformative technologies generate ecosystems.

AI has begun doing the same.

Organisations are redesigning workflows rather than merely adding a chatbot. Software increasingly carries intelligence inside ordinary processes. Jobs are being broken into tasks so that some can be automated while others remain human. Education is being forced to reconsider which abilities are worth testing once producing an answer becomes cheap.

Much of the eventual AI economy may consist of practices that are difficult to imagine while attention remains fixed on whether a chatbot can write a competent email.

A new mathematical toolkit

OpenAI announced in September 2026 that an internal system had produced a proof that three-dimensional fluid flow under the Navier-Stokes equations can develop a singularity in finite time, settling one of the Millennium Prize problems. Mathematics suits machines unusually well, since formalisation allows a computer to check a proposed proof line by line, separating what merely appears true from what is true.

A recent discussion on The World, the Universe and Us and New Scientist (What Does It Mean Now AI Has Solved Navier-Stokes?) offered a sharper way to think about the transition. The link to PodcastThe limitation sits in the why.

Algebra met geometry and caused a commotion

Descartes showed that the two fit together, so that equations could describe geometric shapes exactly. Some mathematicians feared their craft would shift from the beautiful geometric forms inherited from the Greeks to the pushing of symbols across paper, leaving them drone-like drudges with no real understanding. Modern physics followed instead, and every branch of mathematics developed across the next three centuries has drawn on those relationships in some form.

A decade of work, then thirteen minutes

Mathematicians spent more than ten years constructing five fair go-first dice, built so that several players could each roll once with no tie and a clear winner. A science journalist asked ChatGPT for a six-dice version and had a working set thirteen minutes later, produced on the free tier and existing nowhere on the internet. The researchers confirmed it worked, elegance aside.

The prompting was encouragement rather than technique

Plain English throughout, with nothing specialised about it. The model stalled, suggested dice of differing sizes, was told to carry on, and came back with an answer.

Formulating the problem, defining success and explaining why a result matters stay with the human. Once exploration becomes cheap, value moves towards deciding which answers deserve to be found.

The same technology enters different cultural systems

Technological history is often written as a single timeline.

Adoption rarely behaves that way.

The questions a society asks about a new technology are shaped partly by what it already values, trusts and fears.

Printing spread rapidly across much of Europe, while Arabic-script printing developed far more slowly in the Ottoman Empire. The reasons were mixed: political authority, religious institutions, established manuscript and calligraphic traditions, economic interests and the technical difficulty of reproducing Arabic script all played a part.

The machinery existed.

The surrounding system made adoption follow a different path.

More recent examples show the same pattern.

Mobile money moved through Kenya faster than through many wealthier economies because the gap it filled was larger. Cash retained its position in Germany and Japan long after adoption curves elsewhere suggested it would fade. Privacy expectations, institutional trust and established habits mattered alongside technology.

QR-code payments became ordinary in Chinese cities while shoppers in parts of Europe were still getting used to contactless cards.

Capability may be global.

Acceptance is local.

Organisations are discovering something similar with AI.

Give every office the same model, licence and training deck, and adoption can still diverge sharply.

In some teams, experimentation begins before policy arrives. Elsewhere, written permission provides the confidence to begin.

In some workplaces, a junior analyst can challenge an AI-generated answer attached to a senior colleague’s name without much hesitation. In others, hierarchy makes that socially harder.

Some professional cultures treat machine assistance as another tool in the workflow. Others attach authorship closely to individual competence, making disclosure more sensitive.

None of those responses says much about the underlying capability of AI.

They say a great deal about the environment into which that capability arrives.

Language adds another layer. Model performance varies across languages, especially where training data is thinner. A team working primarily in Bengali, Vietnamese or Swahili may judge the same tool very differently from a team working in English.

Both assessments can be accurate within their own conditions.

National culture is only one influence. Profession, organisation, generation, region and individual preference can matter as much or more. A software team in Bengaluru may resemble one in Berlin more closely than a government department in the same city.

The broader principle survives the caveat.

Technology supplies capability.

Culture influences which uses feel natural, legitimate or threatening.

Both sides usually hold part of the truth

Technological optimism can be as simplistic as technological pessimism.

Agriculture brought disease alongside abundance.

Printing spread propaganda alongside knowledge.

Industrialisation generated exploitation alongside extraordinary productivity.

Cars created mobility alongside congestion, pollution and road deaths.

The internet connected billions of people while creating new forms of fraud, surveillance, misinformation and dependency.

The survival of a technology proves little about whether its critics were wrong.

Societies often advance by discovering that both sides held part of the truth.

The technology creates value.

The critics identify costs.

Institutions slowly learn to manage both.

Laws change. Professional standards develop. Education adapts. Social expectations shift. Some behaviours become unacceptable. Others become ordinary.

Eventually the technology becomes embedded so deeply that the original debate begins to look strange.

AI may be different. The reaction is familiar.

Artificial intelligence may prove unusually consequential because of its speed, reach and capacity to replicate cognitive tasks.

It deserves scrutiny.

Questions about employment, intellectual dependence, concentration of power, misinformation, copyright, privacy and autonomous systems are exactly the questions societies should ask while a technology is still taking shape.

Catastrophic risk belongs in that discussion. A low-probability event with irreversible consequences still deserves serious preparation. The lesson from pandemics is not that every warning will prove correct. It is that waiting for certainty can leave institutions unprepared.

Speculative danger can also consume attention that present harms require. Bias, misinformation, energy and water use, copyright disputes and concentrated corporate power are visible now. The podcast used the term criti-hype: magnifying a technology’s most dramatic possibilities while appearing to criticise them. A warning can unintentionally reinforce the claim that the technology is almost limitless.

The sensible position holds both time horizons together. Examine catastrophic risk without allowing it to displace measurable harm. Study present harm without assuming that current capabilities define the upper limit.

History does caution against one prediction in particular:

The world visible during disruption is not necessarily the world that will remain

Socrates could see what writing might take away. He could hardly have anticipated a civilisation organised around accumulated written knowledge.

Early farming communities could see disease, harder labour and unreliable harvests. They could see rather less of the cities, institutions and technologies that settled civilisation would eventually support.

The first automobile was judged against the horse.

The early internet was judged against the newspaper and the shopping mall.

AI is currently being judged against the way humans work today.

That may be the wrong comparison.

AI will almost certainly alter today’s world substantially. The more useful questions concern what humans will build around it, which capabilities will become more valuable, which ones we will preserve deliberately, and which costs we will decide are unacceptable.

Humans have rarely entered technological revolutions gracefully.

We resist. We exaggerate. We underestimate. We regulate late. We adapt unevenly.

Then, quite often, we redesign society around the very thing that once appeared to threaten it.

The habit worth examining may be our confidence that we can already see where the change ends.

A negotiation in Pune

Sixth floor, monsoon outside. An Indian distributor along one side of the table, a Korean components manufacturer along the other, and between them the interpreter.

Volumes agreed, tooling agreed. Then the Indian managing director raised the discount on the second order, and the senior Korean executive said: “Gwaenchanayo”

‘That’s fine,’ the interpreter said.

Twenty minutes later the Indian side offered fifteen extra days on payment. The Korean executive used exactly the same expression. “Gwaenchanayo”
‘No, thank you.’

The managing director put down his pen. ‘Same word, twice. First a yes. Now a no. One of those is you.’

‘Neither, sir. Both are correct.’

One expression, built around gwaenchanta, takes on at least six separate jobs, sorted entirely by the situation. Politeness level alters the ending without altering the range, so the identical expression arrives as (gwaenchanseumnida) in a formal negotiation and (gwaenchanayo) once the room warms up.

The explanation ran ninety seconds, in English, at speed. Across the table the Korean executive watched an animated conversation he had not been invited to, arising from a sentence he had never spoken.

‘Nothing was said. Both of you have been talking for two minutes. Explain, please.’

The interpreter explained, in Korean. The Korean executive found the idea interesting, and his questions took a while.

The managing director listened to two and a half minutes of Korean, none of it preceded by anybody saying anything, and set his pen down again. ‘He said nothing. You said a great deal. Same question as before.’

At which point the man in the middle understood that he was the only person in the room who knew what was happening, and that the position carried no authority whatsoever.

Korean is hardly alone.

Acknowledgement is not agreement

One of the simplest rules in cross-border communication is also one of the most widely ignored. Somebody says okay, right, sure, fine, achha, hai or hǎo, and we hear a positive-sounding word and mentally tick the box.

Very often the speaker has said only: I heard you.

The gap between hearing and agreeing can be worth a great deal of money, and the loss arrives late. A misread token produces no argument in the room. Everybody leaves satisfied, the minutes record agreement, and the disagreement surfaces three weeks later, when a shipment fails to move or an invoice arrives against a discount nobody accepted. By then both sides have documentation supporting incompatible readings of the same afternoon.

None of these readings is a rule. Tone matters, timing matters, and so do relationship, age, hierarchy, region and situation. The dependence on all six is exactly what makes the words dangerous.

Sorted by the job they mostly do, the twelve fall into three loose families.

Reception markers keep the channel open. Japanese hai, Indian achha and Mandarin hǎo often sit here, and the risk they carry is over-reading, since a foreign listener hears agreement in a slot where the speaker offered attention.

Transition markers close a topic and open the next. French voilà, Russian davai, British right and much American use of okay work this way, and the risk is under-reading, since a listener who takes the word as agreement keeps talking after the decision has been made.

Stance markers manage the relationship rather than the information. Korean gwaenchanayo, Canadian sorry, German doubled ja, ja and Brazilian pois é fall here, and they carry the widest range, from reassurance to refusal to mild irritation.

Boundaries between the three blur constantly. A single word crosses families depending on where it lands in the turn, which is the difficulty in miniature.

Three closer looks

The Japanese yes. A Japanese colleague says hai, and many English speakers hear yes. The word often sits much closer to: I am listening.

‘We can ship the equipment by the 15th.’ ‘Hai.’ ‘And commissioning can begin on the 18th.’ ‘Hai.’ ‘So we will mobilise your team immediately.’ Silence.

Nothing went wrong in the translation. The mistake happened one step earlier, when acknowledgement was converted into commitment. Anyone who has sat through a Japanese meeting will recognise a related rhythm, where hai arrives every few seconds throughout a presentation, in a pattern closer to a nod than to a signature.

The Indian achha. Anyone working in India discovers how much labour a single word performs. Six readings sit inside it, separated by delivery alone.

An Indian executive can use it forty times in a meeting without agreeing to a single proposal. Length carries much of the load: a clipped achha tends towards understood, while a stretched one leans towards surprise or doubt, and a doubled one often announces a change of subject. The same executive then switches to English and says ‘okay, okay’. The habit has crossed languages. The words are English, and the system underneath them remains Indian.

English speakers do it to each other. A British executive listens to a presentation and eventually says ‘right’, and the presenter hears correct, please continue. The executive may mean: I have understood enough, we should move on.

An American says ‘sure’, which might carry absolutely, or fine if that’s what you want, or I don’t love this and I won’t fight about it. The dictionary offers no help. The rest of the conversation does.

The smallest words carry the biggest context

English okay makes the point without leaving the language.

‘Are you okay?’ ‘Okay.’ ‘Can you do this?’ ‘Okay.’ ‘The shipment is delayed.’ ‘Okay…’ ‘Okay, let’s move on.’ ‘Okay?!’ One word, six messages, no confusion among people who share the system.

Linguists call these expressions response tokens or discourse markers. Their job is to manage the conversation rather than add information, and their meaning comes from position, tone, repetition and what happens next as much as from the word itself. Patricia Clancy and colleagues showed the pattern by comparing English, Japanese and Mandarin. All three languages solve the same problem, which is showing that the listener is still engaged, and none of them solves it the same way.

Four separate forces push languages towards overloading their cheapest words.

Efficiency. Piantadosi, Tily and Gibson argued that ambiguity is a feature of efficient communication rather than a defect in it, since context already does part of the work. A high-frequency expression needs no separate form for every possible meaning if the situation normally reveals which meaning applies. Zipf made a version of the argument decades earlier, framing it as a bargain between a speaker who wants low effort and a listener who wants easy interpretation. Acknowledgement words are the most frequent, lowest-effort items anybody produces in a working day, which puts them first in the queue for reuse.

Position. A recent volume comparing okay across thirteen languages, including Brazilian Portuguese, French, German, Italian, Japanese, Korean and Mandarin, reached a conclusion with direct operational value: the tone of the token and its placement in the sequence are what create the distinct uses. The word is a slot filler. Meaning arrives from where it sits and how it lands, and both travel badly down a poor connection.

Face. Declining is the costly act in every culture yet studied, and cultures with strong face economies route refusal through a word with positive value. Korean gwaenchanayo, Japanese daijōbu, Brazilian tá bom and British right all perform refusal while sounding agreeable, which leaves both parties a way out. Research on Japanese refusals found three recurring features: avoiding an explicit no, attributing the reason to a third party, and offering a reason that both sides understand to be a formality.

History. Each token descends from a content word that lost its weight. Mandarin dui began as a verb before settling into its role as a particle. Hindi achha traces back to a Sanskrit sense of clear or transparent. Korean gwaenchanta remains a live adjective alongside its use as a token. Ordinary language change, running everywhere, in every period.

The four stack neatly. Efficiency explains the reuse, position explains the range, face explains which meanings cluster together, and history explains where the word came from.

A fifth observation sits underneath all of them. Mark Dingemanse and colleagues, comparing how speakers of unrelated languages signal that they have missed something, found the same short questioning syllable turning up across language families, and concluded that it arrives by convergent evolution rather than shared ancestry. Conversation sets identical engineering problems everywhere. Unrelated languages keep inventing the same small tool.

Inside one linguistic community the whole system stays invisible, because everybody shares the decoding key. Trouble begins when the word crosses a border and the key stays behind.

The token survives the switch to English

The commercial risk concentrates at one point, and it is not the point most people watch.

Interpreted meetings are comparatively safe, since a professional interpreter reads the token in context and delivers the meaning rather than the word. Meetings held in English between non-native speakers are where the losses gather. Speakers carry the habits of the first language into the second, so the token looks English and behaves locally.

An Indian manager on a call says ‘okay, okay’ in the slot where achha would go, marking reception. A German counterpart hears doubled agreement, which in German would be closer to impatience. A Korean colleague writes ‘it’s fine’ in an email, meaning the difficulty stays on the Korean side and needs no discussion. An American reads spare capacity and adds scope. Nobody has mistranslated anything. The words are correct, and the traffic underneath them runs on different rails.

Video calls sharpen the problem. Tone and timing carry a large share of the meaning, and both are the first casualties of compression, latency and a muted room. A half-second delay flattens the pause that would have signalled reluctance. Gallery view removes the glance between two colleagues that would have told you everything.

Five things worth listening for

Before treating a positive-sounding word as agreement, five quick checks are usually enough.

1. Notice the lead-in. A pause, a hedge or a soft preface ahead of the word often signals reluctance that the word itself conceals. ‘Well…okay’ and ‘okay’ are different answers.

2. Look at what followed. Compare ‘okay’ with ‘okay, Raj will send the revised drawing by Thursday’. Names, dates, quantities and actions turn acknowledgement into something closer to commitment.

3. Note where it appeared. A word such as yes, hai, right or okay offered while you are still speaking is usually keeping the conversation moving. It means: I am still with you. Agreement needs better evidence, and agreement rarely arrives in triplicate.

4. Register the delivery. Tone, facial expression, posture and the rhythm of nodding carry a large share of the meaning. Where the channel strips those signals, lower your confidence accordingly rather than assuming the words alone will do.

5. Check whether anything changed. After apparent agreement, look for movement. A date accepted, a price confirmed, an owner named, a next action defined. Where nothing changed, acknowledgement is the likelier reading.

The safest sentence in the room

Written confirmation after an important cross-border conversation settles what the five checks leave open, and the wording repays some care.

‘As agreed…’ quietly assumes the agreement already happened, and a counterpart who reads it as a claim rather than a question will often let it stand rather than open a dispute. Something closer to this works better:

‘My understanding is that we will ship 500 units by 15 October, with commissioning beginning on 18 October. Please correct anything I have captured differently from your understanding.’

It names the commitment, makes disagreement easy, and lets the other side correct the record without losing face. Sending it within the hour matters more than polishing it, since memory of the room fades quickly on both sides. A small difference in wording prevents a very large argument later.

Back in Pune

The managing director spends the afternoon questioning how one Korean expression can carry so many meanings. Then, after a further explanation from the interpreter, he responds: ‘Achha.’

The interpreter pauses. ‘Sir, which achha?’

The room laughs, because everyone suddenly sees the shape of the problem. Ambiguity is far easier to spot in somebody else’s language than in your own. Our own right, sure, okay, achha, hai and gwaenchanayo feel perfectly obvious to us, which may be the most dangerous assumption anyone brings into an international meeting.

The email went out at 4.40, nine lines, all of them numbers. Six were right. Both sides found the other three the next morning, roughly three days earlier than usual, and considerably cheaper.

Where the research thins out

Honesty about the evidence belongs in a piece like this.

The general framework rests on solid ground. Response tokens have been studied for forty years, the comparative work across English, Japanese and Mandarin is careful, and the efficiency argument has been tested well beyond conversation.

Coverage of individual words runs unevenly. English okay has an entire academic volume to itself. Mandarin hǎo and duihave been mapped in detail. Russian and its neighbours are served by a volume tracing a whole family of related particles across Europe. Hindi achha and Korean gwaenchanayo, by contrast, have attracted remarkably little dedicated peer-reviewed attention, given that hundreds of millions of people use them daily and a good deal of commerce passes through them. Most of the available description sits in teaching materials rather than journals.

Anyone reading the table above should treat it as a working guide assembled from research where research exists and from practitioner observation where it does not.

Two notes

On the gap. When Culture Collides with Language, the book behind this article, runs to twelve country chapters, and none of them covers the country’s acknowledgement words. Other linguistic features are missing as well, and a future edition will need to carry them. Additions from native speakers are welcome, since range shifts with region, generation and industry.

On the sources. The reading list below covers the linguistics. Requests for papers on a specific language are welcome, and several of the works carry chapter-length treatments of individual markets.

Picture an ordinary Tuesday a few years from now.

India has done nothing dramatic. It has carried on doing what it has done for years, buying discounted crude where it can find it, declining to sign on to a sanctions framework written in other capitals, and insisting that its foreign policy gets decided in New Delhi. None of that is illegal. All of it is sovereign. Somewhere in Washington or Brussels, a compliance committee, an export-control desk, or a national-security memo concludes that this independence carries a price.

The cost does not take the form of a missile or a blockade. It shows up as a login screen that will not load.

By nine in the morning, employees at a large refiner cannot open their email. By ten, a hospital chain finds its cloud-hosted patient records unreachable and its scheduling system frozen. Court clerks in three High Courts cannot retrieve filings. A state secretariat loses email continuity in the middle of an exchange. The spreadsheet that runs a logistics company’s entire dispatch schedule refuses to open. The AI tools that thousands of developers and analysts had built into their daily work return errors. An app used by thirty million citizens for a government service quietly disappears from a store managed on another continent. Bank back-offices running on a foreign enterprise suite slow to a crawl. Maps stop routing. The cloud console that controls all of it asks for credentials that no longer work.

That scenario is not fiction. It extrapolates from events that have already occurred.

Two warnings, six months apart

The first case came from an unglamorous corner of the technology stack: office email.

In July 2025 the European Union placed Nayara Energy on its eighteenth package of Russia-related sanctions, largely because Russia’s Rosneft holds a stake of just under fifty per cent in the Indian refiner. Microsoft then suspended services to Nayara, reportedly cutting access to Outlook, Teams, and the company’s own hosted data, even though neither US nor Indian law required an American company to enforce an EU designation. Nayara’s licences were paid in full. The company called the move abrupt and unilateral, taken “under the guise of compliance” and without prior notice, and went to the Delhi High Court seeking an injunction to restore access to its own digital infrastructure. The court issued notice, and Microsoft restored services two days later, before the scheduled hearing. During the interval, the refiner reported logistics disruptions and cancelled cargo pickups at its Vadinar port.

Once the specifics are set aside, the lesson stands out. A company that operates in India and serves its energy economy lost access to the systems it runs on for several days, with every licence paid in full, because of a regulatory judgement made on another continent. The dispute ended within days. The underlying weakness remained.

The second case sits at the opposite end of the stack, in frontier artificial intelligence. It is only days old.

On 12 June 2026 Anthropic disabled its two most capable AI models, Claude Fable 5 and Claude Mythos 5, to comply with a US export-control directive that cited national-security authorities. According to reporting in the Wall Street Journal, Commerce Secretary Howard Lutnick had written to Anthropic’s chief executive, Dario Amodei. The order was broad. It told the company to suspend access for any foreign national, inside or outside the United States, including its own non-citizen staff. To comply, Anthropic switched the models off for everyone. They had been publicly available for three days.

Officials raised a cybersecurity concern. The models, on this account, could be pushed into finding software vulnerabilities useful for attacks on banks, government networks and other critical infrastructure, reportedly after agencies learned of a jailbreaking method. Anthropic disputed the directive. It said the vulnerabilities involved were minor and already known to other public models, and that such a standard would freeze frontier AI deployment across the whole industry. The episode followed months of friction between the company and the administration, including a Pentagon decision to label Anthropic a supply-chain risk after it declined certain surveillance and autonomous-weapons uses.

For India the timing was poor. India is reportedly Anthropic’s second-largest market. One large IT services firm had been training fifty thousand employees on the models. Another had recently begun collaborations. Thousands of developers spent the weekend migrating applications, and several startups paused launches that depended on features no longer available. The reaction in India came quickly. Zoho’s founder Sridhar Vembu called it a wake-up call and declared that “globalisation is dead”, urging India to build its own AI and, in the meantime, to lean on smaller Indian and Chinese open-source models. The investor Mohandas Pai went further, calling for an annual AI fund several times the size of India’s existing commitments. Venture investors began advising portfolio companies to reduce their exposure to a handful of foreign frontier providers.

The two cases differ in almost every detail except the one that matters. Nayara lost ordinary office email. India’s developers lost the most advanced AI models on the market. In each case the switch was thrown from abroad, for reasons unrelated to whether the Indian customer had paid or broken any law. Control over that switch sits outside the country at every level of the stack, from the mundane to the cutting edge.

The map of dependence

India’s digital dependence runs in layers. Each layer concentrates in a small number of foreign firms.

At the surface sit the productivity and consumer platforms of daily work: Microsoft 365 and Google Workspace, the leading browser, the two mobile operating systems, the default search engine and map. Beneath them run the enterprise systems that keep the economy moving, such as SAP and Oracle. Lower still is the cloud itself, where a data centre in Mumbai or Chennai can hold Indian data while its control plane, billing relationship and product roadmap stay foreign. The AI layer sits on top of that, and the deepest layers, telecom equipment, semiconductors, app stores and cybersecurity tools, sit underneath everything.

One large country treated the problem seriously about a decade ago, and its approach is worth studying despite its politics. China built domestic substitutes at almost every layer, as the comparison below sets out.

China’s censorship, surveillance and closed internet deserve rejection, and India tends to reject them. The strategic logic underneath still holds. Beijing declined to let foreign platforms become permanent dependencies before domestic alternatives existed.

The risk for India lies less in any single company. Microsoft was not the villain of the Nayara case, and Anthropic clearly preferred a different outcome in its own. The risk lies in concentration. When millions of users, thousands of firms and whole critical sectors rely on the same few platforms, the system gains efficiency and loses resilience at the same time. A single dominant supplier works well until the day it withdraws.

India’s route should differ. It can remain democratic, open, interoperable and globally connected while reducing its exposure. The aim is continuity under pressure rather than isolation.

A familiar pattern of denial

One part of the record should reduce anxiety rather than raise it. The current situation repeats a long-standing feature of independent India’s technological history. On most past occasions when access was denied, India eventually built the capability itself.

The closest parallel to the present AI situation is almost four decades old. During the 1980s India needed high-performance computing for tasks that included weather forecasting. The United States blocked export of a Cray supercomputer, on the grounds that the same machine could model weapons. India responded in 1988 by setting up the Centre for Development of Advanced Computing. By 1991 C-DAC had built the PARAM 8000, an indigenous parallel supercomputer that later grew into a family of machines. A civilian requirement was refused on dual-use suspicion, and the refusal created a domestic capability that had not existed before.

Space technology followed a similar course. India signed an agreement with Russia’s Glavkosmos in 1991 for cryogenic rocket-engine technology to power its satellite launch vehicle. Washington imposed sanctions in 1992 under missile-technology-control rules, and Russia later limited the transfer of know-how under pressure. India argued, with reason, that the engines were meant for civilian satellites used in weather forecasting, telecommunications and resource mapping. It spent close to two decades developing its own cryogenic engine and flew a fully indigenous version successfully in 2014. The denial delayed India’s heavy-lift capability by years while pushing the country to develop it independently.

Navigation produced perhaps the sharpest example. During the 1999 Kargil conflict India sought GPS data for the high-altitude terrain, and the US-controlled system withheld it. Positioning, mapping and location turned out to be strategic infrastructure rather than neutral utilities, available in calm periods and conditional in a crisis. That experience became one of the reasons India built NavIC, its own regional satellite navigation system.

The United States widened the approach after the Pokhran-II tests of 1998, adding dozens of Indian scientific and defence entities to its Entity List and tightening controls on high-performance computers and other dual-use technology. Many of those entities were removed about two years later as the restrictions narrowed. An older case ran in the background. After the 1974 nuclear test, the promised supply of fuel for the Tarapur reactor became tied to supplier-country politics, and India turned to other sources to keep the plant running.

The list runs from uranium to rocket engines to supercomputers to GPS, and now to AI models, cloud accounts and office suites. The specific technology changes each time. The underlying risk stays the same. A capability that a society depends on can be withheld at the moment of greatest need, by firms answerable to a foreign government.

The response has been just as consistent. PARAM exists because a Cray never arrived. India’s cryogenic engine grew out of a blocked transfer. NavIC owes part of its origin to the GPS data withheld at Kargil. Technology denial has worked, oddly, as one of the more reliable spurs to research and development in India’s history. The events of the past year carry that same opportunity alongside the embarrassment.

Where India has built its own

India already holds working sovereign capability in several areas, much of it operating at population scale.

Payments offer the clearest case. The Unified Payments Interface now ranks as the world’s largest real-time retail payments system, handling billions of transactions a month and reducing everyday reliance on foreign card networks. Together with RuPay and the public infrastructure of Aadhaar, DigiLocker and CoWIN, it shows that India can design and run platforms for more than a billion people.

Zoho matters for a different reason. Most of India’s IT services firms build software for foreign clients. Zoho instead owns its full product stack and runs its core workloads on its own servers rather than on a foreign cloud. The company is bootstrapped and profitable, competes with Microsoft and Google on features as well as price, and earns most of its revenue overseas. It shows that India can build finished software products rather than only supply technical labour.

The telecom stack has started to loosen a long-standing import monopoly. BSNL now runs an indigenous 4G stack built by TCS, C-DOT and Tejas Networks across tens of thousands of towers, with an O-RAN-compliant 5G effort and an open-source platform from leading institutes in development. A domestic video-conferencing platform, created through a government challenge, serves several High Courts.

The real weakness lies in adoption rather than capability. India often builds a domestic alternative, then treats the foreign default as inevitable. A platform exists, yet procurement renews the incumbent supplier. Ministerial endorsement of the Indian option has not displaced the habit and network effects that keep users on the global one. The distance between what the country can build and what it actually uses now counts as a security problem rather than a question of market share.

From access to assurance, and the leakage problem

India needs a working doctrine of digital sovereignty rather than a slogan. It can be set out plainly.

First, classify digital systems by sovereign criticality. Email and documents for a small private firm are ordinary business tools, while the same software running a refinery, a port, a power grid, a court, a hospital network or a state administration counts as critical infrastructure and deserves different rules. Second, mandate exitability, so that every critical organisation can export its data, documents and workflows in usable formats within a defined window, and no institution gets trapped in a system it cannot leave. Third, require tested continuity plans: fallback arrangements for email, identity, communication and operational records, whether built on Indian commercial platforms, open-source deployments or government clouds. The test there is practical rather than ideological, namely whether the institution keeps functioning when the primary vendor stops responding. Fourth, give procurement preference to credible domestic alternatives wherever they genuinely meet the bar.

Fifth, separate convenience from criticality. A foreign tool can be a sensible choice for marketing automation and a poor one as the sole operating backbone of a strategic national asset. Sixth, accelerate sovereign AI, meaning compute, models, datasets and evaluation standards strong enough to serve critical needs when foreign access is restricted. Seventh, rehearse failure. India runs military exercises and disaster simulations, and it can run national digital resilience drills in the same spirit, deliberately staging cloud lockouts, API withdrawals, app-store removals and enterprise-software outages before a real crisis does.

On the sixth point, India already has a live debate worth holding in the open. One camp, led by investors such as Mohandas Pai, argues for a national mission several times larger than today’s commitments, running into tens of billions of dollars across funds, credit guarantees, hardware and semiconductors. Another, voiced by Sridhar Vembu, cautions that spending tens of billions is the wrong instinct, and that patient research plus heavy use of open-source models can carry India much of the way. Both positions hold some truth. The point is that the choice has become urgent enough to argue about.

Most proposals for sovereignty skip the next part. Once words like “sovereign”, “swadeshi”, “indigenous” and “strategic” turn into procurement keywords, they attract the wrong applicants. A national effort of this kind tends to draw rent-seekers in its early years. Well-connected firms may wrap a thin Indian shell around foreign software and present it as homegrown. Others may win protected contracts through patronage rather than performance, build subsidised products that never face real competition, then lobby to make the protection permanent. India saw a version of this in earlier decades of protected industry, when closed markets often produced captive monopolies with little reason to improve. Any claim that a sovereignty programme will stay clean from the first day deserves scepticism. Some leakage is likely.

Leakage argues for discipline rather than retreat. Most large strategic projects, from railways to telecom to defence manufacturing, lost something to opportunists in their early stages. The projects that lasted were the ones whose rules gradually reduced that loss. The safeguards here are reasonably well understood. Preference should track performance rather than identity, triggered only when a product clears published benchmarks for security, uptime, interoperability and usability, so that a label alone earns nothing. Open standards and genuine data portability should be mandatory, which keeps the country from trading a foreign lock-in for a domestic one and makes white-labelling easier to detect. Open-source deployments deserve preference where practical, since auditable code is harder to fake and harder to capture quietly. India could adopt Vembu’s suggestion of transparent, independent annual evaluations of national capability across critical technology sectors, so that public claims can be measured against published results. Every form of protection should carry a sunset clause, so that support ends once domestic firms can compete rather than continuing as a permanent subsidy. Procurement should run in the open, with published criteria and audit trails, which leaves less room for patronage.

With that discipline in place, early leakage becomes a cost of building rather than a permanent feature. Without it, India risks swapping foreign dependence for domestic extraction, an outcome arguably worse, since the foreign supplier at least delivered a working product.

Sovereignty is the ability to continue

The test of sovereignty has little to do with using foreign technology in normal conditions, which any country can manage. The real test is whether India keeps operating when conditions change, when a foreign court, a sanctions package, an export-control desk or a company’s compliance team acts for reasons unrelated to India’s interests.

Nayara delivered a warning at the enterprise layer, and Anthropic delivered one at the AI layer. The next case may come from cloud, chips, app stores, operating systems, maps or browsers. It will probably resemble the last two, arriving not as an invasion but as a service that stops responding.

The next phase of Digital India has to shift from access to assurance. Access concerns whether citizens and businesses can use digital tools. Assurance concerns whether the country keeps running once those tools are withdrawn, sanctioned, degraded or attacked. India can keep using Microsoft, Google, Amazon, Apple and Anthropic, and it should keep using the best products available. The change required is to stop treating a paid subscription as a form of sovereign control.

The lesson is short. A capability that cannot be operated under stress is not truly owned. A nation would do well to avoid making any such capability indispensable.

On 28 February 2026 a joint American and Israeli operation killed Ali Khamenei, who had led Iran since 1989, along with members of his family and much of the senior command. The American framing presented the strike as a liberation. Iranians would seize the moment to reclaim their country, and a system built around the Supreme Leader would fall once its head was removed.

The months since have run another way. The government held. An interim council formed within a day, and Khamenei’s son Mojtaba was advanced as successor. Mourning crowds filled the streets. Iran’s retaliation arrived faster and heavier than its careful reply of the previous summer. The war spread across the Gulf, shipping through the Strait of Hormuz fell to almost nothing, and oil rose above a hundred dollars a barrel. The assumption behind the strike, that removing the leader would remove the threat, looks more and more like a misreading.

Most accounts of the crisis reach for two familiar vocabularies. Western strategists talk about deterrence, the calculus of threats and capabilities. Regional commentators talk about power, the contest for primacy across the Middle East. Both capture something real. Both leave out the variable that explains why the same events keep producing opposite reactions. The missing variable is meaning.

Wars do not escalate merely because states possess weapons. They escalate because states assign different meanings to the same actions.

A strike that one side reads as deterrence, another reads as humiliation, and a third reads as the making of a martyr. The weapons are identical. The interpretations are not, and the interpretations decide what happens next.

Three cultures, three starting points

Israeli strategic culture begins with bluntness. The society prizes a manner of speech it calls *dugri*, a borrowed Arabic word for talking straight, and the same directness runs through its security thinking. A state with almost no strategic depth cannot rely on surviving the first blow, so its doctrine favours initiative and preemption, captured in a recurring question: “If the threat is growing, why wait?” Applied to Iran, the disposition generated a threat assessment stated in maximal terms and delivered without softening. That framing shaped how Washington came to define the problem.

American strategic culture reaches instinctively for deterrence and coercion. Force is treated as a message, calibrated to change an opponent’s cost-benefit calculation. Pressure, on this view, produces concessions, and the removal of a hostile leader counts as coercive success. Iran’s weakened position after the June 2025 war, the sanctions, the protests and the damaged network of allies, read in Washington as evidence that the pressure was working and that military means now offered more than diplomacy.

Iranian strategic culture rests on two older foundations. The first is a long tradition of honouring the righteous victim, running from the ancient mourning for the slain prince Siyavash to the central Shia commemoration of Husayn, killed at Karbala in 680 and grieved every year since. A system grounded in that tradition tends to convert the killing of a leader into a martyrdom that strengthens cohesion rather than breaking it. The second foundation was patience. For all his revolutionary language, Khamenei governed with caution, and his responses to direct attack were calibrated to satisfy honour while avoiding a war that might end the system. After the United States killed Qassem Soleimani in 2020, Iran struck back at American bases with enough advance signalling to limit casualties, which let both sides retreat. A patient centre kept Iranian reactions inside predictable bounds.

Three Cultures, Three Interpretations

The collision becomes clearer when the same events are read through each lens at once.

Three Read down the columns, three coherent worldviews appear. Read across the rows, the danger appears. The same act carries a different meaning in each capital, and each meaning licenses a different response.

• Iranian restraint that Washington files under pressure working is filed in Tehran under strategic patience, a reserve of dignity rather than a sign of surrender.

• A strike that Israel records as necessary preemption is recorded in Iran as humiliation demanding redress.

• Leadership losses that the coalition counts as coercive success register in Iran as martyrdom, the most mobilising category its culture contains.

None of the three sides is reading the situation irrationally. Each is reading it correctly by its own grammar, and the grammars do not match.

The Missing Variable: Humiliation

Western strategic literature is built around power, interests, deterrence and capability. It leaves little room for humiliation, and humiliation sits close to the centre of Iranian political memory. The catalogue is long: the 1953 coup that removed the elected prime minister Mohammad Mossadegh with American and British help, decades of sanctions and enforced isolation, the 2020 killing of Soleimani, repeated strikes on Iranian soil, and now the death of the Supreme Leader. Each episode entered the national story not only as a loss of capability but as an injury to standing.

Recent analyses of the June 2025 war have framed its aftermath in exactly these terms, treating the regime’s exposure as a wound to honour as much as a military setback. The distinction matters for what follows. A state that has been humiliated does not calculate the way a state that has merely been deterred does. Its first priority becomes the restoration of dignity, and restoration can be valued above safety.

Culture shapes far more than how people communicate. It shapes how they read status, respect and legitimacy.

The Iranian reply to the February strike supports the reading. It came faster and in greater volume than the measured response of the previous summer, and in its opening days Iran launched a larger share of attacks than across much of the earlier twelve-day war. Analysts have taken the shift as a sign that the command structure, and its old caution, had changed.

That change is the heart of the matter. The strike removed the figure most responsible for keeping Iranian retaliation bounded, and it did so in the name of reducing the Iranian threat.

The patient calculator is gone. Left behind are a martyrdom culture, a deep memory of humiliation, and a dispersed command with less to lose. The combination tends to escalate, not fold. A strike calculated purely in the language of capability can land as an assault on status, respect and legitimacy at once, and a state reacts to that kind of injury with far less restraint than a cost-benefit model predicts.

Built to survive decapitation

The speed of the Iranian reply also has a structural explanation, and it points to a second American misreading. Iran had seen decapitation before, and it had prepared for it. The lesson came from Iraq. The rapid collapse of Saddam Hussein’s heavily centralised state

in 2003, and the wider pattern of American operations against centralised regimes, persuaded Iranian planners that a single command centre was a fatal weakness. The killing of Qassem Soleimani in Baghdad in 2020 drove the point home on Iranian flesh.

The response was a doctrine the Revolutionary Guard calls mosaic defence. Formalised from 2005 and built out by 2008 under General Mohammad Ali Jafari, it broke the Guard into roughly thirty-one provincial commands, each with its own weapons, intelligence, logistics and integrated Basij units, and each holding pre-delegated authority to act without instruction from the centre. The governing assumption was blunt: decapitation is not a decisive blow if the body keeps moving. Authority was pushed down through what Iranian officials describe as successor ladders, with named replacements several ranks deep for every senior post, ready to step up the moment a superior is killed.

The design held under the heaviest test it has faced. With the Supreme Leader and much of the high command gone, provincial units regrouped and fired on their own initiative, which accounts for a reply that arrived within hours rather than days. The same feature carries the danger the article has traced. A force built to keep fighting without its head is, by definition, a force that no head can easily call back. Decentralisation bought survivability at the cost of control. The patient centre that once bounded escalation has been removed, and the machine left in its place was engineered to continue regardless of who gives the order. Several analysts of the doctrine have made the warning explicit: the structure that prevents collapse also removes the off-switch.

Where the readings collide

The American decision applied one culture’s meaning to a situation governed by two others. Decapitation reads as victory inside a framework where the leader is the regime and the regime is the threat. It reads as martyrdom inside a framework built on the memory of Karbala. Preemption reads as prudence to a state with no depth to spare. It reads as humiliation to a state whose political identity is organised around resisting exactly that. The coalition acted on its own grammar and misread the other two, and the gap between intention and reception is where the crisis now lives.

The outcome remains open. The Iranian system may yet fracture under the strain, security forces may defect, and serious analysts disagree about the trajectory. The mechanism, though, is already visible, and it has little to do with the balance of weapons.

The Deepest Insight

Israel fears annihilation.

America fears the loss of its credibility.

Iran fears humiliation.

Each fear is rational within its own cultural framework.

When nations respond to different fears, they misread one another’s actions almost by default.

The most dangerous conflicts are not those in which one side misunderstands the other.

They are those in which every side believes it understands the other perfectly.

Further reading

On Israeli directness as a cultural form, Tamar Katriel’s *Talking Straight: Dugri Speech in Israeli Sabra Culture* (Cambridge University Press, 1986) remains the foundational study.

On the pre-Islamic roots of Iranian mourning and martyrdom, see Ehsan Yarshater’s essay on taʿziyeh and pre-Islamic mourning rites, summarised in the Encyclopaedia Iranica entry on [taʿzia](https://www.iranicaonline.org/articles/tazia/).

On humiliation as a driver of the current conflict, see the Foreign Policy Research Institute, [”Humiliation and Transformation: The Islamic Republic After the 12-Day War”](https://www.fpri.org/article/2025/10/humiliation-and-transformation-the-islamic-republic-after-the-12-day-war/).

On why the system did not collapse, see Al Jazeera, [”Why the Iranian regime did not collapse after Khamenei’s assassination”](https://www.aljazeera.com/opinions/2026/3/17/why-the-iranian-regime-did-not-collapse-after-khameneis-assassination), and the Brookings Institution, [”After the strike: The danger of war in Iran”](https://www.brookings.edu/articles/after-the-strike-the-danger-of-war-in-iran/).

On the mosaic defence doctrine and its consequences, see The Soufan Center, [”Iran’s ‘Mosaic Defense’ Strategy: Decentralization as Resilience Factor”](https://thesoufancenter.org/intelbrief-2026-march-9a/); the Jerusalem Institute for Strategy and Security, [”Iran’s Mosaic Defense Faces Its Real Test”](https://jiss.org.il/en/grinberg-irans-mosaic-defense-faces-its-real-test/); and Radio Free Europe/Radio Liberty, [”With Top Brass Dead, Iran Deploys Decentralized ‘Mosaic’ Strategy”](https://www.rferl.org/a/iran-irgc-israel-us-war/33697690.html).

.

That was the problem.

For more than two decades, my work has lived where business, technology, and culture meet. I started as a mechanical engineer in heavy industry. Construction equipment. Cranes. Trenchless technology. Complex international operations. The work took me through India, the Middle East, Russia, China, and most of Asia Pacific.

For three years I lived in Saudi Arabia, managing multicultural teams in a demanding environment. Later, at a US multinational, I helped open offices in new markets, ran the Middle East and the Indian subcontinent, and eventually took responsibility for Asia Pacific.

On paper this reads like a career in international business.

In practice it was an education in misunderstanding.

Not the dramatic kind. Not the sort that ends up in case studies with obvious villains and clean lessons. The misunderstandings I cared about were smaller, quieter, and more expensive.

A polite sentence that read as evasive.

A direct request that read as rude.

A ‘yes’ that did not mean agreement.

A delayed reply that signalled respect to one colleague and disengagement to another.

A careful explanation that one side called professional and the other called avoidance.

These rarely look like cultural problems at first. They look like slow decisions, weak ownership, missed deadlines, or lack of urgency. Only later do you realise people were not disagreeing about the work. They were operating from different assumptions about how work should be written, framed, challenged, softened, escalated, or concluded.

That realisation became impossible to ignore when I was promoted to Director, Asia Pacific.

My remit suddenly covered multiple countries. China, the Philippines, Thailand, Malaysia, Indonesia, India and more. The Indian subcontinent stayed on my plate. The China and Philippines offices reported into me. American headquarters sat across time zones with its own expectations of speed, clarity, and decision-making.

At first I managed the way many international managers manage. By getting on planes.

Quarterly visits. Factory walks. Corridor conversations. Long dinners after difficult meetings. The real work often happened outside the formal meeting room. A tense issue softened over a meal. A hesitation became visible in someone’s body language. A disagreement surfaced only after the senior person had left. Writing mattered, but it was held up by presence.

Then COVID grounded everything.

The corridor conversations disappeared. The factory floor disappeared. The after-meeting clarification disappeared. The dinner where the difficult subject softened over a glass of wine disappeared.

What remained was a screen and a keyboard.

My day reorganised itself around writing. Early mornings with China, where every exchange carried the weight of face, hierarchy, and relationship. Daytime with India, where context mattered and a decision often needed the conversation around it to make sense. Late evenings with American headquarters, where messages were short, direct, and built for speed. The Philippines, Thailand, Malaysia, Indonesia, and other threaded through the day, each with its own rhythm of respect, caution, warmth, and indirectness.

Within weeks the cracks showed.

A short, polite email from a Japanese reseller partner read in Seattle as evasive. A direct American request read in China as aggressive. A Chinese ‘yes’ meant agreement on Monday and reluctance on Wednesday, and the difference lived not in the word itself but in the sentences around it. Velocity dropped. Trust frayed. Problems that looked operational were often embedded in writing.

So I went back to the books.

Erin Meyer helped explain cultural mapping. Richard Lewis helped explain broad communication patterns. Hofstede’s dimensions sat underneath much of the conversation. Andy Molinsky wrote about adaptation. Craig Storti wrote about crossing cultures. The frameworks were valuable. They helped me understand why a meeting had gone wrong.

They did not help me write the next email.

The academic literature held more answers. Those answers were buried behind paywalls, technical language, and hundreds of pages of theory. A practitioner drafting a sensitive reply at midnight had nowhere simple to go.

So I started keeping notes.

Patterns from my own inbox. The structure my Chinese colleagues used to deliver bad news. The way Filipino colleagues softened a refusal. The signals that suggested a Chinese colleague had a serious objection but would not state it directly. The American instinct to put the conclusion first. The German preference for context, evidence, and precision before commitment. The Indian habit of building meaning through relationship and surrounding explanation before arriving at the point.

At first the notes were only for me.

They helped me decode what I was reading and adjust what I was writing. Over time they became something larger. By 2022, during a four-week sabbatical, I started turning them into a book. By late 2023, the first draft had crossed two thousand pages. Far too long, but complete.

Two more years went into cutting it down.

The personal irritations had to go. The patterns that survived distance stayed. Theory remained only where it served the practitioner. What was left was the practical core. How to read cultural signals in writing. How to write back in a way that lands.

That book became *When Culture Collides with Language: Decode. Adapt. Transcend.*

This Substack starts from the same place.

It is not an academic newsletter, though it will draw on research. It is not a travelogue, though it will move across countries. It is not a collection of etiquette tips, though it will sometimes touch on gifts, greetings, silence, hierarchy, honour, directness, and face.

It is about the written work of crossing cultures.

Here I will write about the things that rarely get taught but often decide whether international work succeeds. Why some cultures expect the writer to be explicit and others expect the reader to infer. Why politeness can look like clarity in one place and aggression in another. Why ‘yes’ is one of the most dangerous words in global business. Why some emails begin with context and others with conclusions. Why silence, delay, hedging, apology, titles, emotion, humour, and even emoji carry different weight across borders.

I will also write what did not fit into the book.

Every book leaves material behind. Some topics were too specific. Some too playful. Some deserved their own space. Honorifics. Emoticons. Digital etiquette. Status anxiety. Apology styles. Feedback rituals. Translation traps. Meeting follow-ups. The quiet politics of ‘just checking in.’

The aim is practical.

If you run cross-cultural teams, report to a foreign headquarters, sell across borders, build international partnerships, or rewrite emails because something feels off, this is for you.

I will not promise easy formulas. Culture does not work that way. What I will offer is patterns you can recognise, questions you can ask, and small adjustments that change how a message lands.

Most global misunderstandings do not begin with bad intent.

They begin with a sentence that travelled badly.