💭 What abundance makes scarce

28 August 2026 · 16 min · 3305 words |  source

Technological revolutions and the predictions they attracted: the ones that failed, the ones that held, the ones nobody saw coming, and what all of it implies for AI. Underneath them all, a pattern: what a technology makes abundant makes something else scarce.
Looking at history is often the best way to predict the future.

Humans have been through many revolutions driven by new technology (for the time). And yet, our patterns of behaviour have remained significantly consistent across ages.

Looking at history, our intrinsic motivations and group dynamics repeat quite frequently. We’re not that different from our ancestors even if our modern times are worlds apart.

Our ancient lizard brain still drives more than we expect…

The following is the transcript of an exploration of that space from a chat session with Claude Opus 4.7, which I thoroughly enjoyed reading. Not necessarily for the predictions themselves (the future is not written yet) but for the “food for thought” they bring. I’ve found myself referring to these patterns quite often in several conversations with people.

Sidenote

This also got me thinking: is the exercise below an early example of an era of “hyper-personalised media” to come? 🤔

The following 3 prompts structured the chat session on 20 April 2026 (reformatted for legibility, adding and fixing links, etc.):

Prompt #1:

List all major titanic technological shifts in the history of mankind, chronologically ordered.

For each, list all the claims of future impact those technologies would have that were popular at the time when the technology was growing.

  • ❌ Start with those that did not pass the test of time and ended up not happening,
  • ✅ then those that did indeed come to fruition,
  • 💥 and those significant impacts that happened but no one predicted at the time.

Major Technological Shifts: Predictions vs Reality

1. Control of Fire (~1M years ago)

Predictions weren’t recorded, but retrospectively:

  • 💥 Cooking enabled smaller guts and larger brains (biological evolution)
  • 💥 Social bonding around hearths shaped language development
  • 💥 Extended active hours beyond daylight

2. Agriculture (~10,000 BCE)

  • ✅ Permanent settlements and food surplus
  • 💥 Dramatic decline in human height and health versus hunter-gatherers
  • 💥 Emergence of social hierarchies, property, and organised warfare
  • 💥 Zoonotic diseases from livestock proximity

3. Writing (~3200 BCE)

  • ❌ Socrates feared it would destroy memory and create “forgetfulness in the learners’ souls”
  • ✅ Preservation of knowledge across generations
  • ✅ Administration of large states
  • 💥 Enabled abstract thought, law codes, and eventually science

4. The Printing Press (~1440)

  • ❌ Would primarily strengthen the Catholic Church by spreading scripture
  • ❌ Elites feared it would cause “information overload” and mental confusion
  • ✅ Mass literacy
  • ✅ Democratisation of knowledge
  • 💥 Triggered the Protestant Reformation (the opposite of Church strengthening)
  • 💥 Standardised national languages and national identities
  • 💥 Enabled the Scientific Revolution

5. Steam Engine & Industrial Revolution (~1760-1840)

  • ❌ Trains at 30 mph would suffocate passengers or cause uteruses to fly out of women
  • ❌ Would liberate workers from toil (instead created factory labour)
  • ✅ Mass production and urbanisation
  • ✅ Radical transformation of transport
  • 💥 Climate change from fossil fuel dependence
  • 💥 Rise of the industrial working class and modern capitalism/socialism

6. Electricity (~1880s)

  • ❌ Would be a dangerous novelty; Edison’s DC would dominate (AC won)
  • ❌ Would eliminate the need for sleep (Edison predicted this)
  • ✅ Universal lighting and power
  • ✅ Transformed domestic life
  • 💥 Enabled 24-hour economies and shift work
  • 💥 Foundation for all subsequent electronics

7. The Telephone (~1876)

  • ❌ “No one would want one in their home” (Western Union internal memo)
  • ❌ Would be used mainly for broadcasting concerts and news
  • ✅ Real-time long-distance communication
  • 💥 Reshaped business organisation and city layouts

8. The Automobile (~1900s)

  • ❌ Would solve urban pollution (horse manure was the crisis of the day)
  • ❌ Would remain a luxury for the wealthy
  • ✅ Mass personal mobility
  • ✅ Suburbs and highway systems
  • 💥 Road fatalities as a leading cause of death
  • 💥 Oil geopolitics and car-dependent urban sprawl

9. Radio & Television (~1920s-1950s)

  • ❌ Radio would end war by letting leaders speak directly to citizens
  • ❌ TV would elevate mass education (Edward R. Murrow’s hope)
  • ✅ Unified national cultures
  • ✅ Real-time news
  • 💥 Propaganda weaponisation (Nazi Germany, Rwanda)
  • 💥 Consumer advertising as economic engine

10. Nuclear Technology (~1945)

  • ❌ “Too cheap to meter” electricity by 1980
  • ❌ Nuclear-powered cars, planes, and household appliances
  • ❌ Peaceful atomic excavation of harbours and canals (Project Plowshare)
  • ✅ Permanent shift in geopolitics and warfare deterrence
  • ✅ Significant baseload electricity in some countries
  • 💥 Long-term nuclear waste problem with no resolution
  • 💥 Mutually Assured Destruction as a stabilising doctrine

11. Space Flight (~1960s)

  • ❌ Moon bases and Mars colonies by 2000
  • ❌ Routine commercial space travel by the 1980s
  • ✅ Satellite communications and GPS
  • ✅ Earth observation and weather forecasting
  • 💥 The “Overview Effect” and modern environmentalism (the Earthrise photo)

12. Personal Computer (~1975-1990)

  • ❌ “No reason anyone would want a computer in their home” (Ken Olsen, DEC, 1977)
  • ❌ The “paperless office”
  • ✅ Computing in every home and workplace
  • ✅ Desktop publishing and software industry
  • 💥 Software eating entire industries
  • 💥 Repetitive strain injuries and sedentary work culture

13. The Internet (~1990s)

  • ❌ Would be a neutral, decentralised equaliser forever
  • ❌ “No greater economic significance than the fax machine” (Paul Krugman, 1998)
  • ❌ End of geography and nation-states
  • ✅ Global instant communication
  • ✅ E-commerce transformation
  • 💥 Surveillance capitalism and concentration into a few platforms
  • 💥 Collapse of traditional journalism business models
  • 💥 Misinformation at scale

14. Smartphones (~2007-)

  • ❌ iPhone would flop (Steve Ballmer, Microsoft CEO)
  • ❌ Would simply be phones with better apps
  • ✅ Always-connected computing
  • ✅ Photography democratisation
  • 💥 Teen mental health crisis (documented by Jonathan Haidt and others)
  • 💥 Gig economy platforms (Uber, Deliveroo)
  • 💥 Attention economy and infinite scroll

15. Artificial Intelligence (~2020s-present)

Currently unfolding, so predictions are in flux:

  • ❌ (likely) AGI by specific near-term dates (claims of 2025-2027 are contested)
  • ❌ (likely) Complete white-collar job elimination within years
  • ✅ (emerging) Transformation of coding, writing, and research workflows
  • ✅ (emerging) Accelerated scientific discovery (AlphaFold, drug discovery)
  • 💥 Unpredictable — by definition we can’t list what nobody predicts

—

A note on methodology: “💥 No one predicted” is always somewhat generous — someone usually did, but they were ignored. And the “❌ failed predictions” list is inevitably skewed toward famously wrong quotes that circulate online; many confident predictions simply faded without being catalogued.

Sources:

Prompt #2:

Is there any pattern you find between the things they were predicted and happened and those that didn’t? Any underlying traits or principles that could help gauge future predictions?

Patterns in Tech Predictions That Fail vs Succeed

Predictions that FAIL (the ❌ column) tend to share these traits:

1. Linear extrapolation of the technology itself
People imagine “more of the same, faster” — nuclear cars, flying cars, moon colonies. They extrapolate the artefact rather than the system around it. Reality bends around economics, not engineering.

2. Underestimating second-order effects
“TV will educate the masses” ignored that attention is a market — whoever pays wins the content. Same mistake repeats with every medium.

3. Utopian or dystopian framing
Extreme predictions (end of war, end of work, end of privacy) almost always miss. Reality is messier and more mundane — the technology gets absorbed into existing human patterns.

4. Assuming the current business model persists
Experts embedded in an industry predict from inside it. Western Union couldn’t see past telegrams; Kodak couldn’t see past film; Nokia couldn’t see past hardware. The incumbent’s frame is the blind spot.

5. Ignoring cost curves and diffusion
“Luxury for the wealthy forever” fails because manufacturing learning curves are relentless. Cars, phones, computers, solar panels all followed this path.

Predictions that SUCCEED (the ✅ column) tend to share these traits:

1. They describe capabilities, not specific products
“Instant global communication” (correct) versus “videophones in every home by 1985” (wrong on timing and form factor, right on essence).

2. They follow the cost curve, not the hype curve
Predictions grounded in “this will get cheaper by X% per year” (Moore’s Law, Wright’s Law, Swanson’s Law for solar) tend to land — often slower than hyped but more thoroughly than expected.

3. They’re humble about timing
Roy Amara’s Law: “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” This is probably the single most reliable heuristic in the history of tech forecasting.

4. They identify what becomes cheap and ask what that unlocks
When computing became nearly free, everything adjacent became software. When bandwidth became nearly free, everything adjacent became streaming. The successful predictors ask: what was previously scarce that is now abundant?

The things NOBODY predicts (the 💥 column) share traits too:

1. Social and psychological consequences
Nobody predicts the loneliness epidemic, the mental health effects, the new social rituals. We’re consistently blind to how technology reshapes us.

2. Business model mutations
Nobody predicted surveillance capitalism, the gig economy, or the attention economy as distinct phenomena. The how it makes money always surprises.

3. Concentration dynamics
Early enthusiasts predict democratisation; they miss that network effects and economies of scale produce concentration. The internet was “decentralised” — then Google, Meta, Amazon happened.

4. Political and geopolitical rearrangement
Oil geopolitics from cars. Nuclear deterrence from bombs. Chip wars from semiconductors. Technology redraws power maps in ways specialists rarely see.

A practical heuristic for gauging future predictions

Ask these questions:

Question 🔴 Red flag 🟢 Green flag
Does it extrapolate the artefact or the capability? Artefact Capability
Does it assume the current business model? Yes No
Is the timeline under 5 years? Usually wrong Less confident
Does it predict utopia or dystopia? Usually wrong Messy middle
Does it engage with cost curves? No Yes
Who benefits from you believing it? Vendor/incumbent Disinterested analyst
Does it account for human behaviour absorbing the tech? No Yes

Applying this to current AI predictions

Using the framework on today’s AI claims:

  • “AGI by 2027” — short timeline, utopian/dystopian framing, vendors benefit from the claim. Treat with scepticism.
  • “AI will make expertise cheaper and more accessible” — capability-level claim, follows a cost curve, doesn’t specify form. More likely to hold in some form.
  • “AI will eliminate all white-collar work by 2030” — artefact-level, extreme, ignores how humans absorb tools. Classic failure pattern.
  • “AI will reshape how knowledge work is organised” — vague enough to be robust, capability-level, doesn’t claim timing. Probably right in ways we can’t yet specify.

The honest position on AI is that the 💥 column is where the real action will be — the things nobody is currently predicting. If history is any guide, the most consequential impacts of AI in 2040 will be things that sound slightly absurd if you said them aloud today.

—

Sources:

Prompt #3:

Explore that absurdity space with some wild ideas, taking into consideration the traits of why nobody predicts them and the green flags heuristics from above, together with the existing knowledge of human behaviours, biases and group dynamics

The Absurdity Space: Wild but Structured AI Predictions

Applying the framework: capability-level, follows cost curves, accounts for human absorption, engages with second-order social/psychological/business-model effects, and embraces concentration dynamics. I’ll mark each with the green-flag trait it leans on.

1. The Rise of “Cognitive Hygiene” as a Public Health Category

[Capability × human absorption × social blind spot]

Just as we developed dental hygiene, sleep hygiene, and digital wellbeing, the 2030s will likely produce “cognitive hygiene” — rituals, products, and medical specialities around not outsourcing too much thinking. Expect:

  • Prescribed “unassisted cognition” hours, like we prescribe exercise
  • Schools that advertise “AI-free learning environments” as a luxury good
  • A generation of adults who can’t recall phone numbers discovering they also can’t structure an argument without assistance — and a counter-movement forms
  • “Cognitive atrophy” becomes a recognised clinical condition by ~2035

Why nobody’s saying it: the loneliness-epidemic pattern. Mental and cognitive consequences always lag the tech by a decade.

2. The Collapse of the Job Interview and the Rise of “Provenance Credentials”

[Business model mutation × concentration dynamics]

When anyone can produce a perfect CV, cover letter, portfolio, and interview performance with AI, the entire signalling apparatus of hiring breaks. The response won’t be “better AI detection” (that’s the artefact-level prediction). It’ll be:

  • Provenance credentials — cryptographically signed records of unassisted work done in monitored environments
  • In-person assessment centres making a roaring comeback as premium infrastructure
  • A new class of “human-verified” professionals commanding wage premiums, like organic food
  • Nepotism and personal networks quietly becoming more important, not less, because trust becomes scarcer

Why nobody’s saying it: it’s a business-model mutation, and it sounds reactionary rather than futuristic.

3. Parasocial Relationships with AI Become the Dominant Form of Relationship for a Significant Minority

[Human absorption × social blind spot]

Not “everyone marries their chatbot” (dystopian framing, fails). Instead:

  • 15-25% of adults under 40 report their “most emotionally supportive relationship” is with an AI by ~2032
  • This is not framed as tragic by those living it — it’s framed as freeing
  • Therapists split into two camps: those who treat it as pathology, those who treat it as legitimate attachment
  • Insurance companies start covering “AI companion continuity” because losing your model of 5 years causes documented grief reactions
  • Governments regulate model deprecation the way they regulate pet euthanasia

Why nobody’s saying it: extreme versions get mocked, so the moderate reality gets dismissed by association. Classic blind spot.

4. The Return of the Village (via Weird Economics)

[Cost curve × concentration dynamics inversion]

If AI makes a single person as productive as a 2020-era team of 10, the minimum viable economic unit shrinks dramatically. This doesn’t produce “everyone’s a solopreneur” (too clean). It produces:

  • Rural and small-town revival as high-earning knowledge workers no longer need cities
  • A strange resurgence of local barter, mutual aid, and physical-presence economies — because the only things that remain scarce are atoms, attention, and embodied presence
  • “Third places” (cafés, clubs, churches) become highly capitalised, as they’re the last places humans gather for non-instrumental reasons
  • Property values invert: dense cities decline in some regions, small towns with good broadband and beautiful landscapes boom

Portugal, incidentally, is well-positioned for this.

Why nobody’s saying it: it contradicts the 200-year urbanisation trend, so it sounds absurd. But the forces that drove urbanisation (proximity to work and markets) are exactly what AI dissolves.

5. Epistemic Tribalism Hardens into Something Resembling Religious Denomination

[Second-order effect × group dynamics]

When every group can generate infinite tailored content reinforcing its worldview, and when “evidence” becomes synthesisable, shared reality further fractures — but not into chaos. Into structure:

  • Stable epistemic tribes with their own trusted AI models, fact-checking infrastructures, and educational content
  • Inter-tribal marriage becomes socially fraught in the way interfaith marriage once was
  • “Which model do you use?” becomes a first-date question signalling values
  • Children raised across tribes develop a new cosmopolitan skill: epistemic code-switching
  • Legal systems struggle because juries no longer share a common factual baseline

Why nobody’s saying it: we frame this as “misinformation crisis” (a problem to solve) rather than “new social structure” (a thing to live with). The framing hides the prediction.

6. The Weaponisation of Boredom

[Human absorption × business model mutation]

When AI can produce infinite personalised entertainment, genuine boredom — the unstimulated mind — becomes the scarcest cognitive state. Consequences:

  • “Boredom retreats” as a luxury product, like silent meditation retreats but without the spiritual framing
  • A generation that has literally never been bored shows measurable deficits in creativity and self-knowledge by their 20s
  • Boredom becomes recognised as necessary for identity formation, and its absence as a developmental harm
  • Schools start teaching “tolerance for understimulation” as a skill
  • A black market for “dumb” devices (flip phones, e-readers without connectivity) thrives

Why nobody’s saying it: boredom has always been framed as a problem to solve. Reframing it as a resource requires inverting a cultural assumption.

7. The Geopolitics of Compute Physically Reshape the World

[Concentration dynamics × geopolitical rearrangement]

Not “US vs China AI race” (the artefact-level prediction). The actual shape:

  • Iceland, Paraguay, Quebec, the Gulf States become disproportionately powerful because they have cheap, clean, abundant electricity
  • Nations with water-rich, cold climates develop “compute sovereignty” doctrines, treating datacenters like oil reserves
  • A new category of war — or at least sabotage — emerges around undersea cables and datacenter infrastructure
  • “Compute refugees” — people who flee jurisdictions where they’ve been algorithmically flagged, because model outputs are portable across borders but reputational scores are not
  • Small nations specialise as “compute havens” the way some became tax havens

Why nobody’s saying it: energy and geography are boring compared to models and chips, so they get underweighted.

8. The Great Re-Ritualisation

[Human absorption × social blind spot]

When work, communication, creativity, and companionship are all mediated by AI, humans respond by aggressively ritualising the domains that remain theirs alone:

  • Hand-written letters become high-status gifts (already starting)
  • Cooking from scratch becomes a marker of seriousness, not frugality
  • Live music, live theatre, live sport command enormous price premiums
  • Weddings, funerals, and coming-of-age rites become longer, more elaborate, more expensive
  • A resurgence of formal dress codes, dinner parties, and social choreography — because effortful presence becomes the rarest resource

Why nobody’s saying it: it looks like regression. But every technological abundance produces a counter-movement of deliberate scarcity (organic food, vinyl records, film photography). AI will produce its own, at civilisation scale.

9. Children Learn to Read and Write Later, Not Earlier

[Human absorption × cost curve inversion]

Counterintuitive but structurally sound: if voice interfaces become universally competent, the economic and social pressure to achieve early literacy collapses for most practical purposes. Consequences:

  • Literacy becomes a specialised skill, like Latin — learned well by some, functionally by most
  • This is catastrophic for deep reading, legal reasoning, and historical knowledge
  • A cultural split emerges between “literate classes” and “oral-AI classes” that maps onto other class divides
  • Efforts to reverse this become a major educational and political battleground by the 2040s

Why nobody’s saying it: literacy is treated as a ratchet that only goes up. But it’s a technology, and technologies can be displaced.

10. The Unpredictable Category Nobody Names

[Meta-prediction]

The single most consequential AI impact of 2045 will be something that, if described today, would sound either:

  • Too mundane to be worth saying (“people will organise their calendars differently”)
  • Too weird to be taken seriously (“a new religion will emerge around model interpretability”)
  • Too adjacent to seem AI-related at all (fertility collapse, housing policy, a specific pandemic response)

This is the 💥 column’s defining property: it hides in plain sight until it’s obvious in hindsight.

The pattern behind the pattern

Notice that most of these predictions share a structure:

The technology creates abundance of X, which makes formerly-ignored Y scarce, which reorganises society around Y.

  • Abundant content → scarce attention → attention economy
  • Abundant assistance → scarce unassisted capability → cognitive hygiene
  • Abundant synthetic media → scarce verified provenance → trust economies
  • Abundant entertainment → scarce boredom → boredom as luxury
  • Abundant virtual presence → scarce physical presence → re-ritualisation

If you want to predict the non-obvious consequences of a technology, don’t ask what it enables. Ask what it makes scarce by making something else abundant. That’s where the surprises live.

—

Sources: