In terms of defining a bubble, he says:

• There is no academic consensus on the nature of a bubble
• He defines it as "a 50% drawdown from the peak equity value that is sustained for at least 5 years"
• Further on he says "Ultimately, it means a phase marked by a rapid escalation in prices and investment, where valuations drift materially away from the underlying prospects and realistic earnings power of the assets involved."
• "Bubbles are impossible to diagnose in real time. Only in retrospect do we know whether exuberance was justified or delusional."

I take issue with the final point, that it is impossible to predict a bubble.  Previous bubbles were all successfully predicted by people living through them! Certainty in forecasting is of course impossible, but to make predictions based on your own judgement is surely reasonable?

The market is a distributed machine for setting prices. We have always used machines like this (or "systems" to be more accurate) for solving complex distributed choice questions. Democracy is another.

When the market works (the "Efficient Market Hypothesis") prices correctly represent the discounted value of future free cash flow. We have a collective delusion problem in mistaking the solution for the problem: "Something is worth whatever someone will pay for it" is a statement about the market being, definitionally, correct. This is a similar error to the common belief that democracy exists because somehow the "people are always right". Newsflash: they are not.

Bubbles are a bug in the market. They represent a period where prices are totally detached from a reasonable assessment of future free cash flow. In retrospect they are obvious to everyone. The weird thing about them is that in advance they are typically also obvious to everyone: but a significant portion of market actors pretend otherwise.

This is because markets routinely demonstrate the features of a Keynesian Beauty Contest. Participants don't value an asset by their genuine assessment of its underlying value (discounted value of future free cash flow) they value it by what they think other people will think it is worth, aka Greater fool theory

Often they are right! It is only when the music stops that we find out who the fools were and who made a fortune on the way up.

So how can you tell whether there is a bubble or not? Well, the market will not be able to tell you, because it's malfunctioning. This is literally how bubbles work.

Anyway, to determine whether there is a bubble or not, Azeem presents five "gauges":

1. Economic Strain: investment as a share of GDP
2. Industry Strain: ratio of capex to revenues
3. Revenue Growth: revenue doubling time in years
4. Valuation Heat: p/e ratio
5. Funding Quality: "composite index capturing funding mix"

It's a decent look at some of the available data, but I don't think any of it helps.

Economic Strain, Industry Strain and Funding Quality are relevant to the size of a bubble, which is obviously important. Small asset bubbles happen all the time (look at meme stocks) but their significance is certainly determined by these sort of factors. These gauges though do not speak to the essential "bubbleness" - whether the assets are indeed overvalued.

Of the others, future revenue growth absolutely will tell you if this is a bubble or not, but we have no future data. Historic revenue growth tells us nothing. Azeem nods to this with:

And this is likely a conservative forecast. Citi estimates that model makers’ revenue will grow 483% in 2025. OpenAI forecasts annualized growth of about 73% to 2030, while analysts like Morgan Stanley estimate this market could be as large as $1 trillion by 2028, equivalent to compound growth of ~122% a year over the period.

Of course mad forecasts are a common feature of bubbles, so this probably is more indicative of a bubble than otherwise. It is not a reassuring data point.

p/e ratio only tells us whether the market is valuing current or future cashflow. If p/e is low then this definitely isn't a bubble (there is present cash flow), but a high p/e just tells us investors are betting on growth. This could be a bubble, but might not be.

So is it a bubble?

A useful framework for analysis in my view is to consider where you stand on the following three concepts:

1. Your theory of cognition
2. Your theory of growth
3. Your theory of profit

If language models are genuinely performing cognition (or a sufficient facsimile thereof) and there is a path to usage growth and AI companies can claim the profit from it, then the current capital outlays are justified.

I don't think any of these is at all self-evident and they all look very shaky.

Theory of cognition

The bull case for language models is that they are doing something close enough to cognition to genuinely replace human activity in a broad range of tasks. This isn't a ridiculous proposition - they routinely do things which previously have only been possible with cognition. If it looks like a duck, swims like a duck, and quacks like a duck, then it probably is a duck.

In this case I do not think it is, in fact, a duck. They certainly show the extent to which language can contribute to cognition (hello Sapir-Worf once again), and cognition and culture are so intertwined the argument about where one ends and one begins might never really be finalised. But I do not think the Transformer Architecture itself does cognition.  We are further along the language model sigmoid curve than many hope.

Theory of growth

Maybe one billion people are already using language models weekly. This is an incredible achievement for a new technology. To justify the build out rate this will need to increase, clearly. The new data centres are there to serve new people, not just existing users who will use them more.

But what is the theory of growth? Why are these people not using language models already? They are trivially accessible and either free or very cheap. Everyone has heard of them. Clearly there has to be a theory that describes why another few billion would start using them.

It seems to be a bet that the chat interface is not the final form for language models - that there will be application use cases which mean they can be adopted for entirely new use cases.

I'm very sceptical about this one too.  SaaS developers are building AI into everything and the response so far is not good. Adoption of AI features in to software like Microsoft 365 has been poor so far. My personal experience, and that of those I've spoken to, is that the addition of AI features into common software packages is poorly received at best.

There are definitely some narrower use cases where the application data is structured in a way which doesn't translate easily into text or images (Miro is a decent example), but I am very sceptical that SaaS developers are going to find another two or three billion users for AI.

Theory of profit

Finally, can the big AI companies claim the profit from growth?

The investor bet seems to be that application level features will produce moats and (ideally) network effects. This means the AI labs can claim and hold users forever. OpenAI recently launched ChatGPT Pulse which is this sort of attempt.

This is even more questionable. Models are converging on capability, including open weights models such as Qwen and Llama. Marginal costs are high. This seems like an obvious market where benefits accrue to consumers eventually, and everyone else gets a commodity share.

To conclude

So if we're along the sigmoid towards converging capability, the market will grow but possibly not exponentially and there's little profit to be claimed, then this really does look like a bubble.

This has led to some pretty wild claims:

They rewrote this story after the tweet, since it was such obvious nonsense, but just to spell it out. To make the economy 10% bigger after 5 years would require 2% growth per year. The OBR is forecasting 1% growth so AI would need to triple the rate of productivity growth, immediately. Hopefully it is obvious that this is totally implausible.

Looks pretty bubbly to me.

The three major trends are:

1. the emergence of public cloud
2. the introduction and eventual dominance of agile software development
3. 
   the domination of the FANGS

All three of these trends have had a fundamental impact on how we think of architecture and the kinds of systems which we build, and I don’t think these are widely understood.

The emergence of public cloud

The impact of public cloud on software architecture has been profound.

All architectural decisions are, fundamentally, economic decisions - they are decisions about how to apply resources to achieve some kind of optimum, in the presence of trade-offs.

Public cloud fundamentally alters the economics of many of these decisions.

In particular:

• Infrastructure-as-Code and continuous deployment reduce the cost of complexity and of scale
• Easy deployment patterns for rich infrastructure components means that new components can be included at low cost
• Serverless systems (i.e. scale-to-zero with metered billing) means that components can be very low cost in initially, and so can be introduced easily

The huge range of services available with these characteristics is fundamentally different from the way systems were designed twenty years ago.

Previously most systems architects would have had a very limited toolbox - a few databases, probably a single preferred application server and a few delivery support items, caches, authentication systems and that kind of thing. Most systems would have been built as monoliths using a two-tier or three-tier architecture not because that was the best option but because it was the only feasible one.

All major public cloud companies offer training and certification in architecture and design. AWS offer Solutions Architect and Application Architect courses and exams, and Microsoft offer a Designing Microsoft Azure Infrastructure Solutions course.

The goal of these courses is awareness - the range of services they offer is so huge, and some of them are so complex, that awareness of them is their greatest problem. They do sometimes provide support in how to assemble complete systems as well - Amazon’s “Well Architected Framework” is notable here - but the bulk of the training is in what they offer and how to use it.

Agile software development

The dominant methodology for software development is now that known as agile. I imagine everyone reading this has some idea what it is. The original Agile Manifesto still puts it better than most. It says that agile favours:

1. Individuals and interactions over processes and tools
2. Working software over comprehensive documentation
3. Customer collaboration over contract negotiation
4. Responding to change over following a plan

This has resulted in most software developers being trained to actively avoid heavyweight processes and documentation.

The result of this is to squeeze architectural activities into the margins - decision-making takes place “near the codeface” and is performed by those involved in implementation.

Agile was a reaction to previous approaches which drew inspiration from other fields of engineering, where detailed planning is the norm. These detailed planning approaches were also the norm in software but were found to have serious drawbacks.

In essence, the economics of software are very different from typical engineering. Software has a basically zero replication cost, so once the software is designed and written it is complete. If cars, for example, could be reproduced at zero cost from the output of the design then (a) cars would be agile and (b) they’d be an awful lot more complicated.

The backlash to document and process heavy design approaches in software was very strong and many developers now have never worked using any approach other than agile.

In my view this has gone too far - there is a sweet spot, which varies from project to project, where both planning and responding to change have value. Accommodating this is difficult, and the emerging approach of Continuous Architecture embodies the attempt to be able to iteratively conduct architectural activities, responding to change but also having a clear plan.

The domination of the FANGs

The “FANGs” are a humorous attempt to describe the mega-tech corporations. FANG stood for Facebook Apple Netflix Google. There have been many other similar acronyms.

These businesses now dominate the global economy and form the underlying infrastructure of our lives. Their decisions have wide-ranging impacts on everything.

In the technology sector they are the place many people want to work, and the way they design and build their products now fills the articles, books, webcasts and conferences that practising developers and engineers consume.

This means the architectural approaches they use dominate our discussions. Many of the practices we consider best practice come from these companies.

A lot of this is very positive, books like Nicole Forsgren’s Accelerate have been hugely influential and provide a lot of underlying data showing how lean techniques can be very effective.

The challenge here is that the vast majority of architecture is not being done at a FANG. Most of it happens inside non-software businesses building systems for themselves, and often those systems just don’t need the characteristics of Netflix.

These systems don’t need infinite scale, or seven nines uptime, and designing them to provide it is unhelpful.

I have talked to many developers who have really gone all-in on some of these approaches, implementing systems using microservice architectures, or using serverless systems, where really they would be fine with existing monolithic two-tier or three-tier designs.

These decisions can have huge economic impacts, increasing development effort substantially, but without generating any real business benefit.

This is not their fault - if everything you read (including from the training provided by public cloud companies!) tells you to do this, then why would you not?

The end result of these trends is that:

1. Architecture is often done late, informally and without structure
2. The dominant approaches used are often inappropriate for the systems being designed
3. Nobody learns to analyse their architectural requirements well, because the typical requirements of the FANGs dominate

Surprisingly, perhaps, there is very little training in architecture provided in courses for Computer Science or Software Engineering, and because of the above trends very few developers and engineers ever really experience architecture being performed in a structured way.

This list skews very heavily to adaptations of science fiction and graphic novels, which obviously is a thing I like, but I think is also somewhere that streaming has really allowed to flourish - these stories often take time to develop and rarely work as movies. They also need the investment in the sets and effects to make them work.

Many of these were great books that would have seemed unfilmable once, so its great to see them work in this format.

The real stars of this list are the following, and I think everyone should give these a go:

• The Expanse
• House of the Dragon
• Severance

Each of these transcends their genre and does something genuinely great.

Anyway in no particular order, these are the series available on streaming platforms in the UK which I recommend. I have tried to link to the actual series themselves, but who knows how stable the links are going to be.

Sky Atlantic

House of the Dragon

A sequel to A Game of Thrones, but genuinely better in my view, with great scripting, casting and acting. They've leaned-in to the "flawed individuals discussing politics in darkened rooms" which was the great strength of GoT, even if the dragons and the boobs were what was promoted.

Currently only one season, but another is in the works.

Amazon Prime

The Peripheral

An excellent adaptation of the William Gibson novel. Currently only one season and no official confirmation of a second, but it looks likely.

The Boys

3 seasons of this adaptation of the Garth Ennis graphic novels, and they get better and better. Antony Starr deserves an Oscar for his portrayal of Homelander.

Reacher

Simple stuff but a well put together adaptation of the book Killing Floor by Lee Child. Only one season so far, but a second one is being made.

Station Eleven

Annoyingly this wasn't picked up by Amazon directly, so this requires an additional subscription to Lionsgate+.

This is another adaptation, of the wonderful novel by Emily St John Mandel.

The Expanse

An adaptation of the series of science-fiction novels by James S.A Corey. I loved the novels and this is a very faithful version that captures much of what was great about the books. This is grown up science fiction, with realistic physics (except where there is Weird Alien Shit going on).

The Man in the High Castle

Initially based on the book by Philip K Dick, this has 4 seasons and develops its own plot. Executive produced by Ridley Scott and with a brilliant performance by Rufus Sewell as John Smith.

Preacher

Flawed adaptation of the graphic novels, which are some of my favourites. I would love to really recommend this, but it really didn't quite work in my view.

American Gods

Based on the book by Neil Gaiman, this has a great performance by Ian McShane.

Apple TV

Severance

Worth getting Apple TV for on its own, this is genuinely excellent in every way. Cast, script, plot, acting, sets, camerawork - the lot. Utterly enthralling and unpredictable. I cannot recommend this enough.

For All Mankind

This didn't get anywhere near the sucess it deserved. Brilliantly scripted and totally unpredictable.

Ted Lasso

Everyone has heard of this. Warm and funny.

Slow Horses

Great adaptation of an excellent book. Second series is in production.

Netflix

Dark

Weird German time-travel series with some great acting. I really loved this. Make sure to watch it in the original German, with subtitles, the dubbing really doesn't help.

The Sandman

Extraordinarily faithful adaptation of the Neil Gaiman graphic novels. Shows that adaptations don't have to change much to work well - I wish they'd adapted Preacher like this.

1899

Created by the same people who made Dark, this is another weird Lovecraftian tale. Make sure to watch this with subtitles too, although the option you want is English this time. There are a lot of different languages spoken and again the dubbing kind of ruins it if you leave it on.

The Witcher

You can't have avoided hearing about this. An adaptation of the novels by Andrzej Sapkowski this is actually really quite faithful. The novels are a bit cheesy. It really took off in the English-speaking world with the release of the excellent computer game Witcher 3.

Henry Caville is great as the lead character, but unfortunately won't be in season 4 due to DC playing footsie with him over his role in Superman (which he now is also not doing). Liam Hemsworth will be taking over, but Henry Caville really made the role his own and I am sceptical if it will work.

The Good Place

Another unavoidable show. The first few series were legitimately great but it kind of lost its way and went on a bit long. Definitely worth watching the first few seasons though.

Locke and Key

Another graphic novel adaptation, this is very faithfully done.

Cyberpunk Edgerunners

Spin off from the massive-but-flawed Cyberpunk 2077 computer game from CD Projekt Red (who made Witcher 3). This is bang on anime and if you like anime it is definitely worth checking out. A single season only, it ties up a single storyline well.

You don't need to know anything about the game to watch it.

Arcane

Another computer game spinoff, this one from Riot's League of Legends. You don't need to know anything about the game (I don't).

The animation for this is really strong, as is the story and the acting. Even if you don't generally like animated series I think you should try this one.

The Silent Sea

Korean science-fiction. Unusual and interesting.

Hellbound

Another Korean series, this is original, shocking, quite weird and highly recommended.

Archive 81

Lovecraftian horror, based on a podcast of the same name. I really loved this a lot, as did loads of other people, but the bastards at Netflix cancelled it after one season - something I am still really cross about.

Norsemen

Norwegian comedy series about a group of Vikings. The cast is norwegian but it is filmed in English. Incredibly dry humour this is brutal and hilarious.

Travelers

There are three seasons of this Canadian sci-fi drama, which I really liked.

These exercises always end up trying to satisfy various different goals - do you recommend a book because it is notable, even if it sucks? How about a book you loved reading but it is never going to change history?

Obviously books which are (a) important, (b) widely recognised as great literature, and (c) you actually enjoyed reading are easy to choose, but that's a pretty small Venn diagram.

So this list has ended up being a mixture of all three categories. For what it is worth I am largely unmoved by whether something is "great literature", and these only make this list if they tick the important or enjoyable boxes too. Recommending something on the grounds it is great literature alone is like recommending a remarkable 1950s mathematics paper. Sure it is constantly referred to and people in the field reckon it's great, but do you really want to spend a week plodding through it? Not personally, no. If you do then all power to you, but I'm not recommending it purely on that basis.

Along with that I am quite sceptical of the "literature vs genre" split anyway. "Literature" is itself is a genre, with common themes and interests. Is this about a middle aged white guy having an affair? It's literature!

Finally this is entirely a personal list and whether I have read something or not is in itself pretty random.

By author

Joseph Conrad

I love Conrad, but I completely understand people who don't. I'd recommend:

• Lord Jim (1899)
• Heart of Darkness (1899)
• Nostromo (1904)

Heart of Darkness is obviously the famous one, since it is the book on which  the movie Apocalypse Now is loosely based, but I think Lord Jim and Nostromo are better.

Franz Kafka

These books are weird and timeless. I'd suggest:

• The Metamorphosis (1915)
• The Trial (1925)

Ernest Hemingway

• A Farewell to Arms (1929)
• For Whom The Bell Tolls (1940) - a great book set during the Spanish Civil War
• The Old Man and the Sea (1952)

Herman Hesse

• Narcissus and Goldmund (1930) - my personal favourite
• The Glass Bead Game (1943) - genuinely quite weird

Jorge Luis Borges

Another personal favourite, actually quite hard to get hold of some English versions of these books. All of them are great, but particularly:

• The Library of Babel (1941)
• Fictions (1944)
• Labyrinths (1962)

George Orwell

• Homage to Catalonia (1938) - more Spanish civil war (and endlessly fascinating conflict)
• 1984 (1948) - so miserable and depressing, but so good

Anthony Burgess

• A Clockwork Orange (1962) - highly notable and brilliantly written
• Earthly Powers (1980) - I loved this, and it contains one of the best opening sentences in all of literature

John Le Carre

I cannot recommend Le Carre highly enough. I don't really get on with his most recent books, but the Smiley series is genuinely a high point of 20th Century literature and A Perfect Spy ought to have won the Nobel Prize for Literature.

• The Spy Who Came In from the Cold (1963) - the third book of the Smiley series, but probably the best one to start with. Once you've read them all, Audible have the fantastic The Complete George Smiley Radio Dramas.
• A Perfect Spy (1986) - genuinely one of the best books of the 20th Century. Criminally under-rated.

Umberto Eco

• The Name of the Rose (1980)
• Foucault's Pendulum (1988)

James Ellroy

• The Black Dahlia (1987) - first of the LA Quartet
• American Tabloid (1995) - first of the Underworld USA series

Tibor Fischer

• Under the Frog (1993)
• The Thought Gang (1994)
• The Collector Collector (1997)

Iain Banks

Really embodies the spirit of Generation X in my view. I loved everything he wrote. He wrote under two names, Iain Banks for his "proper" books and Iain M Banks for his science fiction, because they didn't want readers being confused.

• The Wasp Factory (1984)
• Consider Phlebas (1987) - the first of the Culture novels. Worth reading all of them, very important in science fiction and genuinely great.
• The Crow Road (1992)

Neal Stephenson

• Snow Crash (1992) - seminal cyberpunk. Invented the Metaverse (the good one, not the Facebook one).
• Quicksilver (2003) - the first book of the Baroque cycle, a huge sprawling meander through the enlightenment.
• Anathem (2008) - mind-blowing and also huge and sprawling.
• Seveneves (2015)

China Mieville

One of my favourite authors, you should read everything he has written, but particularly:

• Perdido Street Station (2000) - his first book, and you can tell. The first of a series of science fiction stories, but the themes he develops are common across all his work.
• Embassytown (2011)
• The City & the city (2009)

Subjects and genres

On or about the counterculture movement

• Junkie, William Burroughs (1953)
• On the Road, Jack Kerouac (1957)
• Naked Lunch, William Burroughs (1959)
• Stranger in a Strange Land, Robert Heinlein (1961) - Generally considered science fiction, but this is a unique book in the Heinlein canon. Heinlein mostly wrote right-wing wish fulfilment nonsense which dated terribly. This book is very different (although there is still a bunch of male wish fulfilment) and is genuinely worth reading.
• The Electric Kool-Aid Acid Test, Tom Wolfe (1968)
• Fear and Loathing in Las Vegas, Hunter S Thompson (1971) - his masterpiece, but everything he has written is worth reading
• Zen and the Art of Motorcycle Maintenance: An Inquiry Into Values, Robert Pirsig (1974) - a great introduction to eastern philosophy, blew my mind when I read it. Possibly less mind-blowing if you are already au fait with the subject matter.
• The Dispossessed, Ursula Le Guin (1974) - again normally considered science fiction, but it is really about anarchism
• The Illuminatus! Trilogy, Robert Shea and Robert Anton Wilson (1975+) - hilarious and chaotic. Hail Eris!

Science Fiction and Fantasy

Books from the so called "Golden Age"  and "New Wave" periods of Science Fiction often get recommended, which I think is pretty dumb personally. Many of us read those books as kids and so nostalgically remember them really positively but honestly try reading them now. They've dated awfully.

So just like all that dreadful Victorian literature (I'm looking at you Dickens) that nobody would even publish today, all that stuff can stay in the past unless you have a personal interest.

Fantasy is a genre that probably has more shit in it than any other, with the possible exception of Crime. It is full of sprawling wish-fulfilment epics where a teenager discovers their special power/heritage/god and then goes on to save the world.  Often enjoyable, but very easy to find.

I've tried to pick books here that even if you despise SF&F they are worth reading.

• Lord Foul's Bane, Stephen Donaldson (1977)
• The Gunslinger, Stephen King (1982) - King is a genuinely great writer. The gunslinger books (of which this is the first) are flawed and strange and definitely worth trying
• Neuromancer, William Gibson (1984) - the original cyberpunk. Can be kind of hard going, but seminal.
• Ender's Game, Orson Scott Card (1985) - the author is an awful person but this and it's sequels are brilliant
• Hyperion, Dan Simmons (1989)
• A Fire Upon the Deep, Vernor Vinge (1992)
• Vurt, Jeff Noon (1993)
• Northern Lights, Philip Pullman (1995)
• A Game of Thrones, George R.R Martin (1996)
• Gardens of the Moon, Steven Erikson (1999) - one of the rare fantasy series to make the list, I love all of these
• Revelation Space, Alastair Reynolds (2000) - the first of the Revelation Space books, I loved all of these
• Light, John M Harrison (2002)
• Altered Carbon, Richard Morgan (2002) - the first of the Takeshi Kovacs books
• Jonathan Strange & Mr Norrell, Susanna Clarke (2004) - wonderfully written and realised
• Blindsight, Peter Watts (2006)
• The Name of the Wind, Patrick Rothfuss (2007)
• Leviathan Wakes, James S.A Corey (2011) - the first of the Expanse series of books. Impossible to put down and wonderfully written and plotted.
• Wool, Hugh Howey (2011) - notable (along with the Martian) for being self-published and then being really good and also hugely popular.
• The Martian, Andy Weir (2011)
• Ancillary Justice, Ann Leckie (2013)
• The Long Way to a Small, Angry Planet, Becky Chambers (2014) - notable as a good representative of 21st science fiction. Distinctly non-masculine space opera, this reads and feels completely different from SF of the past.
• Europe in Autumn, Dave Hutchinson (2014)
• Children of Time, Adrian Tchaikovsky (2015)

Other fiction

Last but definitely not least.

• Brave New World, Aldous Huxley (1934)
• I, Claudius Robert Graves (1934) - hugely enjoyable Roman romp about a previously obscure emperor
• Titus Groan, Mervyn Peak (1946) - flawed but unforgettable. The second book in the series, Gormenghast, is even more of the same. The third book is mad.
• East of Eden, John Steinbeck (1952) - moving and grim
• Catch-22, Joseph Heller (1961) - hilarious
• Love in the Time of Cholera, Gabriel Garcia Marquez (1985)
• Mother London, Michael Moorcock (1988)
• Oscar and Lucinda, Peter Carey (1988)
• London Fields, Martin Amis (1989)
• Stone Junction, Jim Dodge (1990)
• The Secret History, Donna Tartt (1992)
• Cocaine Nights, J.G. Ballard (1996)
• Infinite Jest, David Foster Wallace (1996)
• Any Human Heart, William Boyd (2002)
• The Road, Cormac McCarthy (2006) - genuinely disturbing, this gave me actual nightmares
• Wolf Hall, Hilary Mantel (2009)
• Red Plenty, Francis Spufford (2010)
• A Man Lies Dreaming, Lavie Tidhar (2014)
• Station Eleven, Emily St John Mandel (2014)
• Cloud Atlas, David Mitchell (2004)

Non-Fiction

• A Rumor of War, Philip Caputo (1977)
• Godel, Escher, Bach an Eternal Golden Braid, Douglas R Hofstadter (1979) - epic and mind-blowing
• The Man Who Mistook his Wife for a Hat, Oliver Sacks (1985)
• A Short History of Nearly Everything, Bill Bryson (2003)
• The Tiger that Isn't: Seeing Through a World of Numbers, Michael Blastland, Andrew Dilnot (2007)
• The Diving Bell and the Butterfly, Jean-Dominique Bauby (2008)
• The Life of Thomas More, Peter Ackroyd (2008)
• Soccernomics (also called Why England Lose), Simon Kuper & Stefan Szymanski (2009)
• David Attenborough Life Stories, David Attenborough (2009)
• The Second World War, Antony Beevor (2012)
• Strategy: A History, Lawrence Freedman (2013)
• A Rumor of War, Philip Caputo (2014)
• SPQR: A History of Ancient Rome, Mary Beard (2015)
• The Fate of Rome, Kyle Harper (2017)
• Britain's War: Into Battle (1937-1941) & Britain's War: A New World (1942-1947), Daniel Todman (2016, 2020)
• October: The Story of the Russian Revolution, China Mieville (2017)
• Hero of Two Worlds: The Marquis de Lafayette in the Age of Revolution, Mike Duncan (2021)
• The Anglo-Saxons A History of the Beginnings of England: 400-1066, Marc Morris (2021)
• The Dawn of Everything: A New History of Humanity, David Graeber & David Wengrow (2021)

A special shout out to Norman Lewis, who wrote some wonderful travel books in the 50's, and a great memoir:

• A Dragon Apparent: Travels in Cambodia, Lous and Vietnam (1951)
• Golden Earth: Travels in Burma (1952)
• In Sicily (2001)

A word of warning about wrong non-fiction

Lots of the non-fiction I read concerns ancient history or pre-history.

Some of these books claim to have insight into the essential nature of humanity: for example Guns, Germs and Steel by Jared Diamond (1997) and the acclaimed Sapiens by Yuval Noah Harari (2015). These two books, for example, have been hugely influential. They are also despised by experts in their fields.

The problem with this entire genre is the scholarship. They rely on tenuous or frankly non-existent evidence, interpreted heavily through the assumptions of the authors. Many of these books become hugely influential but they really do not stand up very well.

Obviously you should remain sceptical in general when reading any non-fiction. If you are not a legitimate expert in the field then it is easy to be led astray. But I think there is a much greater risk of it right now, particularly in this area. In particular, archaeology has made huge strides in the last twenty years with the advent of new techniques, and many of these books are totally out of date.

I can recommend The Dawn of Everything by David Graeber as an antidote to this. I make no claims as to whether it is actually any more accurate (due to not being the aforementioned legitimate expert), but the authors obviously know their stuff and draw very very different conclusions from, for example, Sapiens, by using actual evidence.

Similarly the end of the Roman Empire looks very different when viewed through modern science in The Fate of Rome by Kyle Harper than when you read even slightly older accounts.

The whole field of history/pre-history/anthropology/political science has not caught up with this new evidence.

A new scientific truth does not triumph by convincing its opponents and making them see the light, but rather because its opponents eventually die and a new generation grows up that is familiar with it...

- Max Planck, 1950

As suggested by "Planck's Principle" above it will take the eventual retirement of a whole generation of academics before there becomes a more settled general view based on this new evidence.

Although you may never of heard of it, this book is something of a cult classic.

Donella Meadows  was an environmental scientist and is probably best known for her book The Limits to Growth, but this book is not strictly about environmental science and is instead about systems more generally. It is out of print, but is available for Kindle, second-hand and there are some PDFs kicking about too.

I first came across this book on a Systems Thinking course hosted by Kent Beck and Jessica Kerr. My personal interest in this comes from the most interesting parts of my work, which is generally on the boundaries between social/organisational, technical and economic systems. In our field we've got various modelling languages for parts of this, such as UML and BPMN, but I wanted something that might be usefully able to work with systems and subsystems of any kind, because often there are complex trade-offs between technical, social, organisational and economic choices. We make these decisions generally through experience and intuition - which means we probably get a lot of them wrong.

Wardley Maps are one option in this space, but they are very high level and my interest is in something with more analytic power at a lower level. I stumbled across "Systems Thinking" as a concept a few years ago, but really didn't get on with the Gerald Weinberg book on the subject. I will perhaps revisit it soon.

System dynamics seems like a genuine option for addressing my need, but learning how to actually model systems in this way is another matter. This book has genuinely useful advice on the subject.

Thinking in Systems really comes in two parts. The first is an introduction to modelling and diagramming System Dynamics (sometimes called differential equation modelling or dynamical systems modelling). The second is a wide ranging discussion of typical behaviours of dynamic systems and their common traps and failures.

System dynamics has its own notation (quite poor in my view), but the concepts, and the notation, are quite simple. As always with any modelling exercise it seems simple until you try and do it - often you discover rapidly that you do not understand the system nearly as well as you thought.

A brief introduction to system dynamics

The rest of this won't make any sense without a very brief introduction.

A system dynamics model consists of:

Stocks - stores of some material. This could be anything you can claim to measure in some way: money, water, people or goodwill.

Flows - movements of material between stocks.

Variables - indicating values or constants that affect flows and stocks

Links - show how other variables, flows and stocks may effect each other

A simple example from the book is this showing how you might choose to drink coffee to give you energy. The boxes are stocks, the taps are flows, the arrows are links and the 'B' indicates a balancing feedback loop. Feedback loops are a critical emergent property of dynamic systems.

The cloud is also a stock but one of infinite capacity. Imagine it like the earth in an electrical diagram.

Creating models

The models in the book are, obviously, on paper, and it is possible and valuable to create them this way - often you may not need the mathematics or graphs to get something useful out of the modelling exercise. Certainly for the models I am interested in just being able to draw them is likely to be sufficient.

However, there are some tools which can help, and you can plug some numbers into them too. One which is probably worth pointing to is Insight Maker, which is free and usable on the web directly. The above coffee drinking system can be modelled pretty easily with this as:

You can then specify some values and simulate the system, producing a graph of the various stocks and variables:

The importance of stocks

The book is full of valuable insights which should be far better understood - particularly amongst economists, politicians and  policymakers, who routinely seem to fail to understand the properties of the systems they are trying to influence.

One of the key insights for me is the importance of Stocks in these models. She says:

Most economic analysis goes one level deeper, to behaviour over time. Econometric models strive to find the statistical links among past trends in income, savings, investment, government spending, interest rates, output, or whatever, often in complicated equations.

These behaviour-based models are more useful than event-based ones, but they still have fundamental problems. First, they typically overemphasize system flows and underemphasize stocks. Economists follow the behaviour of flows, because that's where the interesting variations and most rapid changes in systems show up. Economic news reports on the national production (flow) of goods and services, the GNP, rather than the total physical capital (stock) of the nation's factories and farms and businesses that produce those goods and services. But without seeing how stocks affect their related flows through feedback processes, one cannot understand the dynamics of economic systems or the reasons for their behaviour.

This absolutely rings true for me, and typically when we look at dynamic systems from a high level we don't try to measure or predict stock levels, partly because this is very hard. The processes are easier to see and measure than the absolutes.

She goes on:

Second, and more seriously, in trying to find statistical links that relate flows to each other, econometricians are searching for something that does not exist. There's no reason to expect any flow to bear a stable relationship to any other flow. Flows go up and down, on and off, in all sorts of combinations, in response to stocks, not to other flows.

An example of this that struck me is the infamous Phillips Curve, which is an economic model that relates rates of employment and wage rises. This is exactly the kind of model Meadows is talking about - it ignores any of the underlying "stocks" within the system that actually effect the decisions of the agents within it and just tries to make a gross relationship between the flows. Needless to say, the Phillips Curve doesn't seem to really hold in most cases, but is still routinely referred to.

Something else that struck me about this insight is that computer strategy games, which contain within them dynamic system models, apply stocks rigorously.

For example, a policy might make assumptions about the effect of actions on behaviour - just look at the "nudge unit", but I have never heard of any dynamic systems models of this kind being used (it is quite possible they are of course, just never shared).

Anyway, computer games use this stuff all the time, for example Loyalty in Civilization VI. Your actions have an impact on the goodwill your population feels towards you, and this generates a flow to or from a stock of "loyalty" a city has towards you. As this crosses various thresholds it impacts the way that population behave.

This seem intuitively a realistic model for how people actually behave, and modelling something intangible such as loyalty is genuinely not insane even in a real world model?

Once you start thinking in this way you can see it everywhere, and I have to admit to having become a bit obsessed.

For example, listening to Appendix 3 of Mike Duncan's (excellent) Revolutions podcast, which is a detailed analysis of the progress of a revolution, it is hard not to imagine stocks and flows of intangible collective mental states.

System traps

One section of the book lists common system traps, and how to address them. I won't list all of these, you can just go read the book, but hopefully you will get the flavour.

I recognised all of these from my everyday life working with organisational and technical systems, but also all of them from government policy. That they are so identifiable, and so general, gives me hope that we might get better at fixing them - but also they are ways in which systems naturally fail, so we're likely to keep seeing them forever.

Here are the traps she lists:

• Policy resistance
• The tragedy of the commons
• Drift to low performance
• Escalation
• Success to the successful - competitive exclusion
• Shifting the burden to the Intervenor - Addiction
• Rule beating
• Seeking the wrong goal

By conceiving of systems in simple terms it is possible to describe these in concrete terms which might help designing your way out of them. For example the well-understood problem of the Tragedy of the Commons can be reframed:

When there is a commonly shared resource, every user benefits directly from its use, but shares the costs of its abuse with everyone else. Therefore, there is very weak feedback from the condition of the resource to the decisions of the resource users. The consequence is overuse of the resource, eroding it until it becomes unavailable to anyone.

With possible ways out:

Educate and exhort the users, so they understand the consequences of abusing the resource. And also restore or strengthen the missing feedback link, either by privatizing the resource so each user feels the direct consequences of its abuse or (since many resources cannot be privatized) by regulating the access of all users to the resource.

Hence the constant tension between regulation and privatisation as possible approaches to our increasingly large global commons.

Conclusion

As with all of these analytic/synthetic approaches it can be genuinely difficult to learn and apply, since it requires a change in the way you think about something, when you probably don't have an in-depth grasp of it anyway (or you would already be doing it).

Examples you work with are either toys, which have little transferrable analytic power, or are of complex domains you don't already understand, so either they seem simple and obvious or entirely mysterious.

Learning something of this kind is therefore a process of sort of edging forwards, developing an understanding of the method and of your domain together, and requires quite a lot of work. This is part of why it is so difficult to develop.

Having been through some of this process though, this is actually a great introduction to system dynamics, because of the combination of examples and then higher level discussion of the nature of systems.