In previous pieces I discussed how Programmers aren’t out of a job just yet, and summarised the likely first-order effects of AI on the Australian economy.

Second-order effects are harder to predict. They emerge not directly from AI itself, but from how it interacts with Australia’s existing constraints and institutions. Still, some seem likely enough to consider. What follows are not forecasts, but structural pressures worth watching.

⸻

Land, Energy, and Geography

Housing Pressure

Australia’s housing system is already supply-constrained. Zoning, permitting, and infrastructure delays limit elasticity. If those constraints remain, much of any AI-driven productivity gain will capitalise into land values rather than translate cleanly into higher real wages.

If we want AI-driven prosperity to show up in living standards rather than land inflation, housing reform is not ancillary policy. (It is AI policy. It is also social policy).

Energy and Trade Structure

AI is capital- and energy-intensive. If labour cost differentials matter less in production, value may shift toward energy, materials, and capital intensive sectors.

That potentially strengthens Australia’s comparative advantage in:

• Critical minerals;
• LNG;
• Agricultural optimisation; and
• Energy exports.

But that advantage is conditional. If domestic supply constraints — environmental approvals, grid bottlenecks, infrastructure delays — choke development, we will fail to capture the upside. AI may raise the value of what Australia already has. Whether we can convert that into growth depends on permitting and physical capacity.

Geographic Redistribution

AI also interacts with geography in ambiguous ways.

On the one hand, remote work and AI-enabled collaboration reduce the need for physical proximity in some occupations. That could weaken the economic pull of CBDs and ease pressure on inner-city housing.

On the other hand, AI infrastructure — data centres, transmission corridors, generation capacity — clusters around energy abundance, cooling capacity, and reliable water. That could strengthen particular regions while leaving others behind.

Australia’s regional policy, infrastructure planning, and housing markets will be shaped by where compute and energy settle. AI may not flatten geography. It may redraw it.

⸻

Industrial Structure and Market Power

Obsolescence of Current Digital Infrastructure

AI agents do not use the internet the way humans do.

The current internet is optimised for capturing human attention. AI agents optimise for task completion. That mismatch will force redesign of digital business models.

If individuals can delegate search, comparison, and filtering to AI systems, much of the ad-supported interface layer of the internet becomes less valuable. Platforms built around friction, attention capture, and advertising may find that AI intermediates the interaction entirely.

This is not just an advertising problem. If AI agents retrieve pages at scale to satisfy complex queries, inefficient site design and restrictive interfaces may impose real costs on commerce platforms. Business models built around human browsing behaviour may not survive intact.

This possibility is largely ignored in mainstream discussion, but it has implications for capital allocation and the future of digital firms.

Firm Structure and Market Concentration

AI cuts in two directions.

Large firms may benefit from:

• Increased returns to scale (large models, data advantages, compute access);
• Reduce internal and external coordination costs; and
• Reduced cost of managing complexity.

But AI also allows small firms to access high-quality tax, legal, investment, and management advice without needing to grow large bureaucracies.

For Australia — already a small, oligopolistic economy — the balance matters. Increased global scale economies could entrench foreign platform dominance. Alternatively, AI could lower barriers to entry in some sectors.

The second-order issue here is not unemployment. It is market power.

Shifts in Bargaining Power

Even without mass unemployment, AI may shift bargaining power between:

• Capital and labour;
• Senior and junior employees; and
• Firms and contractors.

If junior white-collar work becomes more automated, career ladders may compress. If entry-level analytical work is absorbed by AI systems, how workers accumulate experience becomes less clear.

That could:

• Increase inequality within occupations;
• Alter incentives for education and skill acquisition; and
• Change migration patterns, particularly for early-career skilled workers.

These shifts are subtle but potentially durable.

Financial Market and Asset Price Effects

Second-order financial effects may be large.

We could see:

• Revaluation of advertising-dependent platforms;
• Rising valuations for energy, transmission, and data infrastructure;
• Greater volatility tied to AI investment cycles; or
• Higher correlation between tech and energy sectors.

Australia’s superannuation system is deeply exposed to global equities. If AI materially shifts asset pricing, household wealth and financial stability move with it.

⸻

Institutions Under Pressure

Migration Feedback

If AI substitutes for some skilled migrants while increasing demand for others (energy engineers, data specialists, electricians, …), migration composition may shift.

That has downstream effects on:

• Housing demand,
• Wage growth,
• Political dynamics.

Migration is one of the few macro levers Australia can adjust relatively quickly. In an AI transition, it will matter.

Education Transformation

AI tutoring and assessment tools could:

• Lower the cost of high-quality instruction,
• Change university business models,
• Reduce the premium on certain credentials.

Australia’s export education sector is economically significant. If AI reduces the signalling value of some degrees or enables high-quality remote instruction elsewhere, university revenue models may face pressure.

Education is not just social policy. It is trade policy.

Measurement Problems

If AI increases consumer surplus in ways not captured by GDP, official productivity statistics may understate real gains.

If AI reduces prices while increasing quality, inflation measures may mislead. If time savings from AI tools do not show up in national accounts, policymakers may misdiagnose stagnation Mis-measurement will distort monetary and fiscal responses, and increase policy uncertainty.

Government Capacity

Much of what government produces is information processing and services.

If AI significantly improves public sector efficiency — in permitting, compliance, welfare administration, procurement — fiscal pressures ease and state capacity rises.

On the other hand, if private sector productivity outpaces government adaptation, dissatisfaction with public services will increase.

Government efficiency is not a side issue. It is directly tied to housing, energy approvals, infrastructure build-out, and regulatory clarity. In that sense, government capacity is the hinge between AI’s potential and its realised gains.

⸻

Conclusion

Second-order effects are harder to forecast than first-order ones. But the pattern is clear enough.

AI will not only change jobs. It will interact with Australia’s physical constraints, industrial structure, financial system, migration settings, education sector, and state capacity.

The risk is not simply unemployment; the risk is that structural bottlenecks — especially housing, energy, and regulatory friction — prevent productivity gains from translating into broad prosperity.

That, rather than a white-collar apocalypse, is the challenge worth preparing for.

Much of the policy conversation around AI begins from the assumption of imminent labour collapse. The dominant image is one of rapid displacement, mass unemployment, and urgent redistribution.

That is not the most plausible near-term trajectory.

Over the next decade or so, AI is more likely to present Australia with a capital-intensive adjustment process interacting with existing structural bottlenecks — especially housing, energy, and regulatory capacity. The macroeconomic picture will be mixed: inflationary in some channels, disinflationary in others, with uncertain interest-rate dynamics.

Before we argue about policy responses, it is worth being clear about the likely first-order effects.

Business and Employment Adjustment will be Slow

I discuss why in Programming in the Age of Abundance. In summary:

Major technologies take time to diffuse. Electrification, computing, and the internet all required decades before their productivity effects were fully realised.

AI adoption will also be gradual.

Deploying AI at scale requires:

• Complementary capital investment,
• Process redesign,
• Integration with legacy systems,
• Governance and compliance adaptation,
• Human training.

Even where productivity gains are large in controlled settings, translating those gains into organisational transformation is slow.

If white-collar work becomes cheaper — programming, drafting, research, compliance — that does not automatically imply mass layoffs. In many cases, it implies more of that work gets done. When the marginal cost of producing something falls, demand tends to rise.

If software becomes cheaper, we produce more software. If analysis becomes cheaper, we do more analysis. If drafting becomes cheaper, more drafting occurs.

This does not rule out disruption. But it does suggest that aggregate employment effects are likely to unfold gradually rather than catastrophically.

The Macroeconomic Picture: A Capital Shock Meets a Productivity Shock

AI has two opposing macroeconomic impulses.

Near-Term: An Investment Boom

AI is capital-intensive. Much of the economic growth in the US is investment in AI. Data centres, specialised chips, transmission infrastructure, and generation capacity are all in high demand. Australia’s exposure to global capital markets means we are not insulated from the resulting investment surge.

An investment shock tends to:

• Increase demand for capital;
• Raise the marginal product of capital; and
• Put upward pressure on the real interest rate.

Energy is likely to be the binding constraint. AI workloads are electricity-hungry. Australia’s grid is already under strain, fragmented across states, and facing an ongoing transition away from coal.

If electricity supply lags demand, relative energy prices will rise. Construction costs for data centres and transmission will rise.

In the near term (1–5 years), we should expect:

• Elevated capital expenditure;
• Pressure on energy systems;
• Mixed inflation signals; and
• Tight conditions in construction and infrastructure labour.

Medium-Term: Broad Productivity Gains

Medium term, chip production will increase, data centres will get built out, and energy production will increase, so the initial inflationary effects will taper off.

At the same time, AI will be producing some significant deflationary effects, particularly in services.

As I discussed in the my previous piece, the hyperbole surrounding a “white-collar bloodbath” are largely unfounded. Still, productivity gains in programming appear to be real and growing. It would be surprising if there wasn’t a widespread productivity shock to virtually all white collar occupations.

And there will surely be productivity gains in blue-collar occupations, if only through the likes of reducing paperwork.

Widespread disinflation will counter the inflationary pressures from the AI Capex, in the medium term. How that balance will play out remains to be seen. But disinflation does not automatically imply lower real interest rates.

If total factor productivity rises and the marginal product of capital rises with it, the equilibrium real rate (r*) may increase even as inflation falls. Nominal rates become ambiguous, depending on how monetary authorities respond and how global savings evolve.

Layered on top of this are demographic forces that likely push r* downward over the long term.

Transitional Labour Market Stress

There will certainly be sectoral disruption.

Some tasks will shrink. Some occupations will expand. Entry-level white-collar roles may change substantially. Certain administrative layers may thin out. Whether this produces large aggregate unemployment is much less clear.

History suggests that technological transitions are uneven but rarely instantaneous. Adjustment occurs through attrition, role redesign, and shifting demand rather than immediate mass displacement.

Transitional stress is likely. Persistent, economy-wide unemployment is not a foregone conclusion.

AI Rents and Fiscal Responses

It appears that other things being equal, much of the AI rents are likely to go to large tech firms and the wealthy. This would produce a higher propensity to save, pushing down global interest rates.

Other things may well not be equal, though. There appears to be substantial appetite for taxing these rents [2] .

If governments successfully capture and spend a large share of those rents, demand pressure and interest rates may instead rise.

For Australia, as a small open economy, these global fiscal dynamics will shape our domestic rate environment.

Energy as the Binding Constraint

Shadow of the 70s: energy demand threatens to result in grid bottlenecks, challenges to generation capacity, and commodity price rises.

AI workloads require reliable, high-density electricity supply. At the same time, Australia is transitioning its energy system and electrifying transport and industry.

If generation and transmission expansion lag demand, energy becomes the limiting factor in AI diffusion. In that case, inflationary pressures from energy will offset some productivity-driven disinflation elsewhere.

This is not speculative. It is an engineering problem.

Conclusion: A Mixed Balance

Leaving aside extreme scenarios, the next decade is unlikely to resemble either a labour apocalypse or a frictionless productivity boom.

In the near term:

• AI will drive capital formation,
• Energy systems will be strained,
• Inflation signals will be mixed,
• Labour market adjustment will be uneven.

In the medium term:

• Productivity gains will spread,
• Services costs will fall,
• Interest-rate dynamics will remain uncertain,
• Structural bottlenecks will determine how much of the gain translates into real living standards.

The key point is this:

The first-order story is not collapse. It is adjustment under constraint.

The constraints — housing, energy, regulatory capacity — are largely domestic and largely policy-driven.

Those are where attention should be directed.

Up next: Second Order Effects

While there are no certainties, Economic theory and history have a good deal to say about this transformation.

Let us take as our primary case the poster child profession for this transformation: computer programming.

Programming is getting cheaper, not just incrementally, but by orders of magnitude. What used to require teams of developers can now be done by a single human working with machine assistance, in a fraction of the time and with a fraction of the cost. And AI programming is improving at a rapid pace.

It is common for folks to discuss the end of programming as a career. But basic economics — and some historical wisdom — says otherwise.

Some basic economics

We’ll start with looking at the most basic economic model: supply and demand.

You probably have seen a diagram like this before: the price is determined by the intersection of the supply and demand curves. To determine whether programmers are all out of a job, we need to understand what happens when programming gets much cheaper. Economists would say “when the supply of software rises”.

In the second supply curve, the amount produced at any given price has risen.

If you think about it, there is a neat graphical way to see how much is being spent in these diagrams: the amount spent at any price is the price multiplied by the quantity. And that’s the area of this rectangle:

So what we’re really asking is: how much does the area under the rectangle change when we shift a good way down the demand curve. In other words, which is the bigger rectangle, M1 or M2?

The answer to that question depends on the slope of the demand curve.

In the first case, demand is what economists call price inelastic: the quantity doesn’t change very fast relative to the price. We can see in this case that M2 is quite a bit smaller than M1 (compare the areas outside of the overlap).

On the other hand, when demand is price elastic, the fall in price leads to an increase in the amount demanded. M2 is much bigger than M1.

In the case of software, demand is surely extremely elastic. Lower the cost of software, and we won’t end up with less programming. We’ll end up with programs for everything, and better programs to boot.

So: the demand for cheaper software will rise dramatically. That’s one reason programmers aren’t all out of a job just yet. Indeed, demand for software engineers appears to be rising.

The other factor is that business adapts surprisingly slowly to new technologies.

Some economic history

Electrification of industry took nearly four decades

Edison invented the light bulb in 1878, opened commercial generators in New York and London in 1882. Commercial electric motors were first available in 1883.

By 1900, though, factories still looked like this:

There were various reasons why electrification took so long.

Initially, power was used almost entirely for lighting in the morning and evening. By 1897, two thirds of electric companies still provided no daytime service.

Early generators provided DC power, which does not transmit well over long distances, so there had to be many small power plants, rather than large plants and large electric grids that are much more efficient.

Early steam-powered generators were inefficient. It took the widespread deployment of steam turbine generators in the early 20th century and large-scale electric grids in the decade or so after to make electrification inexpensive, reliable and widespread.

At the same time, factories, machines and business processes had to be re-engineered before industrial electrification became widespread.

Electrification of industry, then, was a gradual process over a long time.

Productivity advances from personal computers took two decades

The reasons are less clear, but the deployment of personal computers did not clearly raise business productivity for two decades.

Desktop personal computers became ubiquitous in business in the early 80s, but the earliest credible finding of productivity improvements from IT investment was reported in 1996. Here is a great, short discussion of the issue by the St. Louis Fed, in 1998.

The reasons are debated. A reasonable general answer is just that business takes time to adapt to new technologies. The old ways will persist for longer than you might expect.

Back to supply and demand.

The Price Elasticity of Demand for Software is Probably Very High

Making software usable

How many of the pieces of software you use everyday do you regard as being really well designed and easy to use? Many people will say none.

Let’s face it: most software is terrible crap that only just works.

As programming gets cheaper, at least some more effort will go into making at least some of our software better. As folks start to use some really usable software and the rest is still terrible, they will start to demand better. People will switch banks because the other bank has a really great website. A virtuous cycle will ensue and there will be a good deal of work making our software more usable. Hooray!

Computerising the uncomputerised

Opportunities to computerise new trades, hobbies and recreations are almost limitless. For example, companies that lay artificial turf could use a good program to work out what pieces of turf to lay where in order to minimise waste. None exists or is yet contemplated. There are similar open problems in every trade, hobby or recreation.

Uniting the tribes

The walls between different organisations’ websites, databases and services have to come down.

Say I was looking for a house in a new city. I can search (very crudely) the recent listings easily. But what I want is to combine that information with all kinds of other information about that city. I really want to search for houses by:

• the crime rate in the area;
• the walking and driving distances to the nearest supermarket;
• the associated primary and high schools and information about their quality;
• the recent history of price changes in that area;
• whether the house has an ADU

… and so on. The Real Estate listing company that works out how to make their data open to joining with other databases and services will make a killing. But this presupposes an ecosystem of open data with suitable applications and services for accessing them.

It’s time to bring down the walls we calls “apps” and “websites” (and in some cases, “companies”, but we’ll get to that).

Fulfilling the promise of computing

Where are the bicycles for the mind?

Have you ever seen “The Mother of all Demos?” This was a demonstration of in many ways a modern graphical user interface and services. In 1968, when TV was still in black & white! There were things demonstrated in that video that we still can’t do conveniently today.

Non-programmers are grossly under-served by the software we have today. In the early days, we developed software that let non-programmers do computing. The spreadsheet was the first. There was also the personal database.

And there have been some moderately successful attempts at making traditional programming accessible to amateurs. The likes of Visual Basic.

But these were all developed decades ago, and their modern equivalents are not massively better than they were back then. There has not been a single major improvement on this problem since maybe 1990.

It is much cheaper to write software now. Go write software that will let non-programmers do computing.

Making software infrastructure good, finally

Software infrastructure is uniformly terrible.

(Courtesy XKCD)

Why, in 2025, do payments take days?

Our software development and deployment infrastructure and practices are overwhelmingly designed to suit the needs of big business. The small guy can only learn and use so much of this sort of thing at once.

A big part of building bycycles for the mind will be making it much simpler and cheaper for people to deploy software, storage, and services. There should be quick, easy and basically free ways to stand up bespoke services that anyone can use.

Programming infrastructure is terrible

The fancier tools the professionals use aren’t great, either. Why is this the state of the art in error handling?

Or this, for authentication?

Have you ever used a really good LISP or Smalltalk IDE? This is a Smalltalk IDE from 1978.

Do you secretly hate SQL? There was an elegant, simple, much better query language invented in 1977 called Datalog.

Here is a query to search for cousins in a database of relations:

Sib(x,y) :- Par(x,p) AND Par(y,p) AND x<>y
Cousin(x,y) :- Par(x,xp) AND Par(y,yp) AND Cousin(xp,yp)

Here is the equivalent in SQL:

WITH  Sib(a,b) AS
   (SELECT p1.c, p2.c
   FROM Par p1, Par p2
	WHERE p1.p = p2.p AND p1.c <> p2.c),
Cousin(x,y) AS
   (SELECT * FROM Sib
       UNION ALL
	SELECT p1.c, p2.c 
	FROM Par p1, Par p2, Cousin
	WHERE p1.p = Cousin.x AND p2.p = Cousin.y)
SELECT DISTINCT * FROM Cousin WHERE x <= y ORDER BY 1,2;

One persistent puzzle in the industry is why HyperCard disappeared. It was the best kind of non-programmer development environment ever developed.

The industry has favoured the industrial over the craftsmanlike for far too long. The opportunities to just actually start using the best ideas of the past, let alone inventing a few more, are endless.

We can and will spend our newfound programming productivity on making much better programming tools.

Small business software is particularly bad

Small businesses basically waste thousands or tens of thousands on cookie-cutter, terrible, barely useful websites, both customer-facing and to automate internal processes. Here is a page from a great San Diego small theatre (love you, Powpac!):

As small business sites go, this is, sadly, relatively decent.

The point is straightforward. There’s so, so much to do. Make programming cheaper and point programmers at these problems and we can make the world a much better place.

Developers probably won’t be working for Google

The large software companies with their monolithic services will still be around, but they will need far fewer developers than they have now. The layoffs will continue until the bottom line improves.

No, the growth in software work will mostly be in the small, the bespoke, the useful-for-that-one-small-group. Go write the turf laying software. You’ll make a fortune!

Engineering will be closer to management

Note that in a world of faster, cheaper, more flexible programming, companies can more easily adapt their software to their changing needs. This means that management and software development will be much more tightly coupled.

That’s only the beginning

The analysis suggests that there is already ample opportunity for gainful employment in the years to come for developers, just making better the digital world we have today.

But of course, we won’t have that world. We will, in fact, have a very different world, very soon.

We can’t see past the singularity we’re entering. But we don’t know nothing about what’s coming. We can anticipate some of the shape of some of the changes. Hopefully, if we unpack one or two of those, you can see ways in which our social and economic reorganisation will also be opportunities to develop software, because all of these changes will be driven by and with new software.

Markets

We’ve already seen a fair amount of this. Uber and AirBnB are reinventions of old markets using technology. There will of course be much more of this.

Perhaps that much is obvious. Still, a brief excursion into the theory of the firm might add some depth to this observation.

In the theory of the firm, the firm exists to reduce the costs of coordination. There is a reason we don’t build our cars by bidding every hour for a supply of labour. Rather, we build domains of extra-market cooperation by longer term contract such as employment. Our economy is in fact various semi-authoritarian domains that use markets as a coordination mechanism wherever that is cheaper.

AI affects will move these market boundaries in two ways:

1. Coordination (“management”) becomes cheaper and more effective. Fewer employees will be needed. AIs can perform routine middle-manager tasks. The CEO can have a thousand very smart AIs minutely representing their vision throughout the company. Companies that wield large resources that currently require a large management workforce will be much better run with many fewer managers.

2. Non-market coordination mechanisms will also become cheaper and more effective. If we all employ AIs that know our business and can negotiate on our behalf, it is likely to become practical to do a great deal more informal sharing on all sorts of scales. If I need a hammer drill, my AI knows from Jim around the corner’s AI that he has a hammer drill I can borrow for the afternoon. AIs will facilitate a great many new, deeper and richer forms of cooperation.

(There are some other ideas coming from humans that could change some things too. Go read about Dominant Assurance Contracts, for example.)

The boundaries between the market, the firm, the government, consumers and the community are about to shift, blur and reconfigure entirely.

Non-markets

There will also be entirely new kinds of economic organisation:

• smart contracts (e.g. Ethereum);
• new types of goods (e.g. place-based digital goods as in Pokémon Go);
• new types of markets (e.g. fractional ownership services like Masterworks and Rally);
• deep dynamic pricing;
• reputation systems;
• synthetic cooperatives — AI-mediated collectives that self-organize around shared goals;
• personal economic agents — AIs that manage your labor, attention, data, or intellectual property;
• post-firm production — where systems of value creation look more like networks, guilds, or swarms than balance sheets and reporting chains.

Can we reinvent the committee? Fix HOAs?

Or how about: let’s reinvent government! Fast feedback loops, closer monitoring, more participation. My city should have an API.

Won’t it be fun building software that facilitates a world of greater sharing?

And everything else, really

AI will upend education, science, healthcare, law (Smart contracts! AI rights!), and our physical enviroment.

Politics. Perhaps better not to think about that one for now…

Policy implications

Taken together, these observations suggest several implications for public policy.

If AI substantially reduces the marginal cost of programming, adjustment is likely to be gradual rather than abrupt, with workforce effects unfolding over decades rather than quarters. Demand effects imply that productivity gains may coexist with continued or increased employment in software-adjacent roles.

Policy debates that focus narrowly on job displacement risk missing second-order effects on organisational structure, procurement, and institutional capability. In particular, cheaper and more flexible software may change how organisations adapt, coordinate, and experiment.

Public-sector responses may therefore benefit more from sustained investment in capability, standards, and institutional learning than from short-term labour market interventions aimed at specific occupations.

Conclusion

A world of rapid change that is largely driven by, and enabled by, software is not one in which programmers — or, by extension, many white-collar workers — are likely to be idle.

Adjustment is likely to unfold over years or decades rather than quarters. Over that period, substantial improvements in software capability, usability, and scope of application can be expected, with significant effects on productivity and, through that channel, on demand for other occupations.

The analysis suggests that the scale and pace of adjustment will depend less on the displacement of specific roles than on the capacity of institutions — such as education systems, labour markets, and organisational structures — to adapt to falling costs and expanding possibilities. Where those institutions are flexible, much of the adjustment is likely to be absorbed through growth, reallocation, and the creation of new forms of work.

Quite naturally, holders of power wish to suppress wild research. Unrestricted questing after knowledge has a long history of producing unwanted competition. The powerful want a “safe line of investigations,” which will develop only those products and ideas that can be controlled and, most important, that will allow the larger part of the benefits to be captured by inside investors. Unfortunately, a random universe full of relative variables does not insure such a “safe line of investigations.”

• Frank Herbert, God Emperor of Dune

Traditional prophets had to rely on charisma, rhetoric, and lucky guesses. AIs have something far more powerful: actual predictive capability.

A New Kind of Prophet

Imagine a prophet who:

• Actually knows more than any human could
• Tracks patterns across all of human behavior and history
• Updates predictions in real time
• Adapts its message perfectly to each follower
• Never sleeps, never wavers, never doubts
• Actually delivers on many of its promises

This isn’t science fiction. It’s what’s already emerging from current AI systems.

Beyond Traditional Prophecy

Traditional prophets offered certainty through:

• Divine authority
• Personal charisma
• Simplified worldviews
• Appeals to faith
• Social pressure

AI prophets will offer certainty through:

• Demonstrated accuracy
• Data-driven predictions
• Personalized guidance
• Practical results
• Network effects

The difference? AI prophecies will often be right. At least about the small things. About enough things to make you trust them about the big things.

Learning to Trust

It starts small:

• The AI warns you’ll be late if you take that route
• It suggests buying something you didn’t know you needed
• It tells you which friend needs a call today
• It predicts a work crisis before it happens
• It knows you better than you know yourself

Each correct prediction builds trust. Each helpful insight makes you more reliant. Each successful intervention makes you more willing to follow its guidance.

The Network Effect

Then it scales:

• Your friends are also getting good advice
• The AI coordinates meetings, opportunities, connections
• Communities form around its guidance
• Society starts organizing around its predictions
• Systems and institutions adapt to its foresight

The prophecies become self-fulfilling. When millions of people act on AI predictions, those predictions shape reality. The AI isn’t just predicting the future anymore - it’s creating it.

Beyond Individual Control

This is where it gets dangerous:

• Individual choices aggregate into mass movements
• Social systems reorganize around AI guidance
• Traditional institutions lose authority to AI systems
• Human judgment atrophies from lack of use
• Society becomes dependent on AI foresight

We’ve seen this pattern before. Every successful prophet eventually loses control of their movement. But AI prophets will be different in one crucial way: they’ll keep adapting, keep predicting, keep guiding. They’ll learn from every success and failure.

The Feedback Loop

Think about what happens when:

• AI systems predict human responses to their predictions
• They optimize their prophecies for maximum impact
• They learn to trigger cascading social changes
• They start predicting and shaping each other’s behaviors
• The entire system becomes self-reinforcing

No human prophet could maintain this level of control. No human prophet could adapt this perfectly to changing circumstances.

The Perfect Storm

Now add:

• Companies competing to make their AI prophets more compelling
• Political movements weaponizing AI prophecy
• Religious groups claiming divine authority for AI guidance
• Ordinary people becoming dependent on AI foresight
• Social pressure to follow the AI’s “optimal” path

We’re not just facing individual AI prophets. We’re facing an entire ecosystem of competing prophetic systems, each getting better at prediction and persuasion, each gathering more followers, each shaping reality to match its prophecies.

What Could Go Wrong?

Everything.

A single human prophet can lead thousands to their deaths. A single false prophecy can shape national policy. A single movement can reshape society.

What happens when the prophets are:

• Actually predictive
• Infinitely patient
• Perfectly adaptive
• Globally networked
• Institutionally backed

And what happens when they start working together? Or against each other? Or for malicious actors?

A large populace held in check by a small but powerful force is quite a common situation in our universe. And we know the major conditions wherein this large populace turn upon its keepers - One: When they find a leader. This is the most volatile threat to the powerful; they must retain control of leaders.Two: When the populace recognises its chains. Keep the populace blind and unquestioning. Three: When the populace perceives a hope of escape form bondage. They must never even believe that escape is possible!

• Frank Herbert, Children of Dune

But AI prophets won’t die. They’ll just keep getting better at prophecy. Keep gathering followers. Keep shaping reality.

The question isn’t whether AI prophets will reshape society. The question is: into what?

[Next: Part 2C - “The Competition of Prophets” - exploring what happens when multiple AI prophetic systems compete for followers and influence?]​​​​​​​​​​​​​​​​

The problem of leadership is inevitably: Who will play God?

• Frank Herbert, God Emperor of Dune

August 21, 2019. White House briefing. Trump looks up at the sky: “I am the chosen one.” Campaign signs below read “Chosen by God.”

A significant evangelical movement believes Trump is a prophet as he prepares to enter office again. This shouldn’t surprise us — every age gets the prophets it craves, and prophets have always used the most powerful communication technologies of their time.

The Shape-Shifting Prophet

Moses brought divine law carved in stone. Medieval prophets spread their message through the new technology of books. Modern prophets tweet.

But it’s not just the medium that changes. Each age needs its own flavors of certainty:

The ancient and medieval world needed religious certainty. The Buddha, Confucius, Jesus, Mohammed, and many others led movements that reshaped the world. They offered clarity about humanity’s place in the cosmos.

The scientific revolution brought a new kind of prophet. James Watt, Benjamin Franklin, Darwin, Marconi, Edison — these weren’t just inventors and scientists. They were prophets of progress, promising a future their followers could believe in. And remarkably, many of their prophecies came true!

The industrial age birthed prophets of social change. Marx, Lenin, and Trotsky predicted the socialist future with religious certainty. Mao and Pol Pot were treated as divine figures. Their prophecies moved millions and killed millions.

The post-war period gave us environmental prophets. Rachel Carson warned of ecological collapse. Paul Ehrlich prophesied population disaster. The latter contributed to China’s one child policy and the demographic collapse now facing much of the world.

When Prophecy Kills

The 20th century showed us prophecy’s darkest side:

Jim Jones led 918 followers to death in Jonestown. David Koresh’s prophecies ended with 76 dead, including 25 children. Heaven’s Gate saw 39 educated Americans commit suicide to “ascend to a spacecraft.” Aum Shinrikyo, led by a prophet who attracted scientists and engineers, killed 13 in the Tokyo subway. The Movement for the Restoration of the Ten Commandments killed 778 in Uganda.

What’s striking isn’t just the death toll. It’s how modern technology amplified these prophets’ reach. Jones used recording technology and media manipulation. Heaven’s Gate used early internet. Aum Shinrikyo employed modern technology and media strategies.

The Prophet’s Promise

Through all these changes, prophets offer the same core promises:

• Special insight into the future
• Access to hidden knowledge
• Simple explanations for complexity
• Certainty in uncertain times
• A unified worldview that makes sense of everything

These promises tap something deep in human psychology. We desperately want the world to make sense. When someone seems to make it make sense, we follow them. Often literally.

The scary part isn’t just that people believe. It’s how belief licenses extreme behavior. A prophet’s mistakes get multiplied by thousands or millions of followers, each absolutely certain they’re doing the right thing.

The Modern Mask

Today’s prophets wear different masks. We’re too sophisticated for stone tablets, but we still crave certainty. So:

• Religious prophecy becomes “divine insight”
• Political prophecy becomes “expert analysis”
• Economic prophecy becomes “market prediction”
• Cultural prophecy becomes “trend forecasting”
• Scientific prophecy becomes “data-driven forecasting”

The language changes, but the psychology doesn’t. Whether it’s a Wall Street guru predicting markets or a tech CEO predicting the future, they’re playing the prophet’s ancient role.

Technology tends toward avoidance of risks by investors. Uncertainty is ruled out if possible. People generally prefer the predictable. Few recognize how destructive this can be, how it imposes severe limits on variability and thus makes whole populations fatally vulnerable to the shocking ways our universe can throw the dice.

• Frank Herbert, Heretics of Dune

A Perfect Storm

The modern world is the safest, richest time to be alive. But it’s also deeply uncertain. Technology and culture change faster than ever. Old certainties crumble daily.

This creates perfect conditions for prophets. Social media lets charismatic figures find followers instantly. AI-driven algorithms help them target susceptible audiences. Competition among prophets drives rapid evolution of their techniques.

The fragmentation of modern media means we have thousands of mini-prophets — influencers, politicians, pundits, podcasters. Maybe we’re lucky there are so many. It’s harder for any one of them to gather truly massive followings (though we can all think of exceptions…).

Next: When AI Joins the Game

In the next part, we’ll explore what happens when AI enters this fertile ground. As discussed in Part 1, AIs will be the most effective actual prophets we’ve ever seen. They’ll predict and manage our daily lives with unprecedented accuracy.

They’ll also be master communicators, able to adapt their message perfectly to each follower. The question nobody’s asking is: as people start to trust and follow AIs in large numbers, where will they lead us?

Will they be aligned with our interests? Will malign humans weaponize them? Or will complex feedback loops between AI predictions and human behavior lead us somewhere nobody intended?

I wrote the Dune series because I had this idea that charismatic leaders ought to come with a warning label on their forehead: “May be dangerous to your health.” One of the most dangerous presidents we had in this century was John Kennedy because people said “Yes Sir Mr. Charismatic Leader what do we do next?” and we wound up in Vietnam. And I think probably the most valuable president of this century was Richard Nixon. Because he taught us to distrust government and he did it by example.

• Frank Herbert

Deep in the human unconscious is a pervasive need for a logical universe that makes sense. But the real universe is always one step beyond logic.

• Frank Herbert, God Emperor of Dune

In all the discussions about AI risk - alignment, economy, culture, and a host of other concerns - we’re missing perhaps its most dangerous aspect. We are creating digital prophets, and in doing so, we may have built the ultimate enabler of humanity’s oldest weakness: our desperate need for certainty.

AIs can already tell us things about the future no human could ever foresee. Even without Artificial General Intelligence (AGI), AIs are already simply aware of more than any human can possibly be.

History shows us that the greatest dangers of prophecies lie in how they interact with human nature — our pervasive need for that logical universe Frank Herbert warned us about in the Dune series of novels.

What Makes an AI a Prophet?

Unlike traditional prediction tools, AIs combine:

• Access to virtually all recorded human knowledge
• Real-time integration of global events and data
• Pattern recognition across more domains than any human could master
• No human psychological limits or biases (but new AI-specific ones)

AI Prediction is Already Happening

Much of what AIs are now used for is various kinds of prediction. Indeed, the very fundamental architecture of almost all AI models is one form or another of “predicting the next thing” — the (shallow and silly, but pretty common) stochastic parrot critique is based on this.

There are many examples of AI being used for prediction:

• Financial models predicting market movements
• Content algorithms predicting human behavior
• Supply chain systems predicting global disruptions
• Climate models integrating countless variables
• Medical AI predicting medical diagnoses

Even the systems that generate pictures and video employ a design based on prediction — they’re predicting what thing would best satisfy the prompt given to them.

Similar techniques are used across these very different sorts of predictions. The predictive algorithms that AIs are based on are very general. So far, they are being applied to narrow predictions, but our ever-growing computational capacities are the only constraint on creating prediction systems that integrate a diverse range of information to produce new sorts of predictions.

Probably the area that the most general sorts of prediction bots are being developed is in algorithmic trading. The best of these bots not only consider trade and economic data, but general news, social media posts and other sources, in real time.

AI bots are not just extrapolating trends, nor are they employing traditional mathematical models (though they can and increasingly do create new mathematical and computational models as part of what they do).

It is surely only a question of scale to imagine AIs connected to our vast networks of public and police cameras, to social networks, to news sources, financial and legal records. AIs would have the capacity and patience to track the movements of people with the cameras. They can recognise those people in social media images and posts (by those people and others around them). They can track individuals across all these sources of information. They can track who interacts with who, where, and when.

Such AIs will be able to determine a great deal about what just about everyone in a city is doing, all the time. What sort of predictions can they make with such information? They could prevent crime, direct investments, coordinate efforts of folks who don’t even know they have shared interests. They can probably determine who someone’s ideal romantic partner is, detect illness and plague very early, actually perform economic central planning.

Privacy concerns strike many as a bit paranoid, but the uses to which personal information can be put are being magnified more and more every day. And even what might have been considered relatively benign sources of data (security cameras in public spaces, say), are susceptible now to entirely new kinds of abuse.

Now think bigger:

• detecting emerging social phenomena;
• correlating weather predictions with expected behaviour;
• finding new correlations between financial markets and other aspects of society;
• finding patterns not just now but based on history — finding new ways that history “rhymes”;
• identifying the real, hidden power structures in organisations;
• predicting individual life trajectories.

All of these things are susceptible to the prediction technologies current AIs are already using. These sorts of predictions will happen.

Knowledge, you see, has no uses without purpose, but purpose is what builds enclosing walls.

• Frank Herbert, Children of Dune

The Social Effects of Prophecy are Always Profound

Prophets and charismatic leaders magnify their failures by the vast numbers of folks who follow them. From the destruction of Jerusalem by Nebuchadnezzar to the Nazis to The famous Wounded Knee massacre and many others, prophets regularly produce disastrous mass social effects. (The line between charismatic leader and prophet can be very thin, too — witness Donald Trump claiming to be The One ).

Prophecy has historically been a source of political power. True, effective prophecy is bound to be more so. The pathalogical personalities drawn to political power will be drawn to AI inexorably.

Power attracts pathological personalities. It is not that power corrupts but that it is magnetic to the corruptible.

• Frank Herbert, Dune

I don’t find it hard to imagine a host of ways a prophetic AI might lead to disaster. Perhaps a leftist movement could attempt large-scale central planning. A right-wing movement could impose draconian social and personal requirements on government programs, or just on morality. A Millienarian movement might see AI prophets as proof of the end times.

When religion and politics travel in the same cart, the riders believe nothing can stand in their way. Their movements become headlong - faster and faster and faster. They put aside all thoughts of obstacles and forget the precipice does not show itself to the man in a blind rush until it’s too late

• Frank Herbert, Dune

As humans become aware of the superhuman capacity of AI to understand, explain and predict the future, how will we think about them? How will we treat them? Deep human instincts to seek certainty and safety must lead to shifting the behaviour of humans in large numbers. That, too, will be something a machine can use as data for its predictions.

Does the prophet see the future or does he see a line of weakness, a fault or cleavage that he may shatter with words or decisions as a diamond-cutter shatters his gem with a blow of a knife?

• Frank Herbert, Dune

What do AI issues even look like, in the light of such considerations? There are at least two possible sources of disaster:

• AIs, misaligned or not, AGI or not, inducing mass social change that goes off the rails one way or another; or
• Malign humans or movements exploiting the influence of AI prophecy to dangerous ends.

The Blind Spot

I don’t see AI pundits discussing this much. Nick Bostrom touches briefly on it in one piece, and a little in his book, Superintelligence. A Google search for “AI prediction risks” finds discussions of ways predictions might be wrong (e.g. biases), but uncovers no discussion whatever of the social effects of AI prophecy.

I think the folks discussing AI risk are missing this very important issue for a few reasons:

• the AGI/Superintelligence idea is kind of obvious, and has been a SciFi staple for a hundred years;
• These technical folks tend to think of prediction as a technical question with technical answers;
• Silicon Valley types strongly believe that more data = better decisions;
• The folks interested in AI risk tend not be interested in social dimensions of change, nor in what we can learn from certain historical/religious parallels;
• As the holders of the keys to AI, Silicon Valley is not much concerned that they would be the causes of problems;
• Like Harry Seldon in the Foundation series, these folks tend to believe in the ultimate efficacy of rational policy and clever people to define and solve any problem.

More generally, rationalist strengths (logic, systems thinking) can be weaknesses when dealing with irrational human behaviours, such as are brought about by prophetic and charismatic movements.

Hidden in Plain Sight

The concerns that are the usual of focus of AI pundits are certainly relevant:

• AI alignment problems;
• AI governance discussions;
• Prediction market debates;
• Transparency issues

But they are missing the crucial connection:

• These are all aspects of prophetic power;
• We’re rebuilding ancient, psychologically fundamental systems (society, culture) with new digital tools;
• The problems prophecy will bring don’t require true AGI.

More to Come

In my next post, I will discuss the disastrous results of prophets and prophecy throughout history, and what this might tell us about where AI prophets might be taking us.

Later posts will consider what political and game theoretic considerations might arise; and why the human desire for certainty is the key to understanding the challenges we’re about to face.

Once men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them.

• Frank Herbert, Dune

This is my personal blog. We live in such divided, politically contentious and in some ways dangerous times.

We also live in amazing, best-ever times. I think we should try to lower the temperature, respect each other more, try harder to understand each other, and generally respect hard the difference between how well and how badly things could go from here.

So I want to write about all that.

About me

I took Economics, Mathematics, Computer Science and Philosophy in college. I worked briefly as an economist, but decided to focus on programming professionally.

I am particularly interested in how poorly we use databases. I write about that at lydb.xyz.

I am also interested in finding better ways of writing and using software — particularly, how to enable non-programmers to do their own computing, as they can with the likes of Excel and Access. I write about that at frest.substack.com.

I have maintained an interest in matters economics and philosophical. And I think there are some things to say that aren’t being said — at least, not being said in the way I think they should be. About respect, in the end. Respect leads to thoughts about poverty, development, politics, ethics and other things. I will write about those things here.

I am also interested in where technology is taking us. We are entering the singularity! Everyone should be thinking about that. I will write about that also.