This week Marc Andreessen sat down on the Founders podcast and said two things:

The first: “I aim for zero introspection. Move forward. Go.”

The second: that Elon Musk is “maybe the greatest manager of our era.”

For people with even minimal intelligence, these two things (no introspection and Musk’s celebrated management) are clearly related. And Marc’s second claim just so happens to be what this paper is about.

In short: the Silicon Valley management gospel (cult?) has a specific epistemology: first-principles thinking beats domain expertise, and the founder who acts without overthinking will outperform the institution that pauses to evaluate. Andreessen actually said the seldom-discussed part out loud: introspection is the enemy; don’t question the model. Just… go forth, wreck havoc, and let someone else deal with the fallout later.

The problem with Silicon Valley’s epistemology, though, is that it is correct in exactly one environment, and catastrophically wrong in every other.

Why? Well…

The infamous first-principles thinking is a cognitive tool optimized for complicated systems: things with many parts but knowable, stable relationships. Like, for example, a rocket engine or a battery cell. Or anything where physics stays constant while you think about it.

But most of the systems that matter, like manufacturing at scale, platform governance, government, healthcare, regulation? These are are complex: their relationships are dynamic, emergent, and change in response to your intervention. You cannot break these things down without destroying the thing you are trying to understand.

This paper introduces a formal model, called the correction window, that explains why the same Elon Musk management playbook produces self-correction at SpaceX, brand erosion at Tesla, false vindication at Twitter, and institutional collapse through DOGE.

As you’ll see, the variable throughout these examples is not leadership quality, but in fact the feedback loop and architecture of the system being managed.

Every organization has two clocks that matter:

1. The first measures how long it takes to find out a decision was wrong.

2. The second measures how long before the damage from that decision becomes permanent.

At SpaceX, the first clock is fast: a rocket fails, and you know within days. And the second clock is slow: engineering knowledge is well-documented, you can rehire, you can rebuild a team. There’s a wide window between discovering the mistake and losing the ability to fix it.

At the US federal government, though, the clocks are reversed. It takes years to discover that firing mine safety inspectors or drug approval staff was a mistake. But the institutional knowledge those people carried disperses within months: they find new jobs, the relationships dissolve, the system reorganizes around their absence. By the time the damage shows up, the window to fix it has already closed.

The gap between those two clocks is the correction window. When it’s wide, you can afford to move fast and break things. When it’s narrow or negative (when the damage becomes permanent before you even detect it) the same approach doesn’t produce bold leadership but irreversible institutional failure.

The federal government lost 317,000 workers in 2025, while spending went up by $248 billion. In this case, thanks to the Musk Management Ideology, the correction window was negative before the first person was fired!

To conclude, Andreessen’s “zero introspection” is not a personality quirk. Very sadly, it is the anti-intellectual foundation of a management method that has now been exported from rockets to government. The paper makes the case that this export is not bold; it is a structural error with a formal mathematical proof.

Read the full paper here →

This article draws on two working papers: “Institutions as Coordination Architectures: Adaptive Bandwidth” and “Market Formation as a Systems Engineering Problem.” Both develop the formal models and cross-domain evidence summarized here. Available on request: s@sinead.co

Subscribe now

If you have a mortgage in Ireland, it means that you either:

(1) Are incredibly (un)lucky that you have been able to buy one of the few, overpriced houses on the island, and (2) that your Irish mortgage is probably a “tracker mortgage”, in which case you already know what I’m about to write in your bones.

Between 2022 and 2023, the ECB raised interest rates by 4.5 % :

• An Irish household on a tracker mortgage saw their monthly repayment rise by over €600.

• A German household on a fixed-rate mortgage paid around €35 more.

Same central bank but radically different outcomes to households.

Most commentary on this framed it as an Irish problem; insert something about mortgage culture, boom-era recklessness, the failure to fix themselves. But that framing is wrong, because this is not just an Irish issue!

Last year, I did a deep dive into the Irish Central Bank’s institutional architecture, in terms of how it’s structured, what it can and can’t do, and where the real constraints on Irish monetary policy actually sit. What came out of that work was a much bigger question:

Why do some EU households absorb ECB rate decisions immediately, while others barely feel them at all?

The answer is structural, and incredibly important (even if a little boring).

Consider that Irish tracker mortgages transmit ECB rate decisions almost perfectly, meaning that Irish people feel the impact of interest rate changes with near-zero lag.

German fixed-rate mortgages, on the other hand, are sitting on some sort of bond infrastructure that dates to Prussian land finance law of 1769 (I know), which insulate households almost entirely.

So, while neither Germany nor Ireland chose this, both inherited the systems that are either very lucky, or very unlucky, depending on where you live.

And the unlucky part? It’s because, if you’re in Ireland, you’re actually redistributive. Meaning that Irish households pick up the cost of inflation such that other countries, like Germany, don’t have to.

In other words: Irish and Spanish households doing more demand-destruction work per basis point is part of what allows the ECB to stop raising rates sooner.

Indeed. German households get the benefit of lower inflation and a stable currency without their mortgage bills having moved! While the ECB’s work was done disproportionately by households in Dublin and Madrid. So while the reward of fighting inflation is shared across the EU, the cost is not.

This structural subsidy (from high-transmission to low-transmission economies) cannot be resolved by market forces, or by anything the Central Bank of Ireland has the power to do.

This paper explains why, and what it would take to fix it.

Read the full paper here →

The frameworks applied here, of architecture lag, premature markets, coordination architectures, are developed formally in two companion papers by Sinéad O’Sullivan. Available on request: s@sinead.co

Subscribe now

The NATO 5% spending target has consumed much attention since Trump appeared for Round Two of his Presidential Era. Given that pressure, Europe is spending more on defense than at any point since the Cold War.

But when the first real test came to deliver one million artillery shells to Ukraine by March 2024, Europe managed 520,000.

Turns out, Europe doesn’t have the industrial architecture to convert money into output. Ooops…

And that gap is the actual European defense problem, not the NATO 5% shenanigans.

In short, NATO’s GDP target measures financial inputs, not capability outputs. The assumption embedded in it is that if you spend enough money, military capabilities will follow. But this has *checks notes precisely zero empirical basis and substantial evidence.

You see, defense (or any!) capability emerges when three architectures co-evolve: technical (weapons systems, interoperability standards), industrial (production capacity, supply chains), and institutional (procurement governance, command integration). You cannot pour unlimited capital into an industrial market that doesn’t exist. (Or, you can, but this is egregiously bad fiscal policy!!)

So, the controversy:

Spain is being called a free-rider for refusing to commit to 5%, and actually last weekend at a dinner party this came up in conversation with my American counter-intel pal. Which means it’s both topical and a sore point.

But Spain's argument that more spending "would only reinforce our dependence" is… architecturally correct. For under the current procurement framework, hitting a higher target means buying faster, and buying faster means buying American.

So Spain's objection isn't to spending more, it's to spending more through a system that routes the money out of Europe. In short, that NATO GDP metric says nothing about where the money goes. And in its current system, the industrial dynamics mean that increased spending does the opposite of what Europeans need it to do (and what Trump says he wants it to do!).

So, today: 60-70% of European defense procurement currently flows to American contractors because European producers can’t deliver at scale, which is a result of fragmented procurement across 27 national systems. And this fragmentation persists because there’s no institutional architecture for European defense market integration.

The spending surge? It’s flowing straight into this trap, and creating another generational-cycle of dependency on American industrial policy.

The two papers below lay out the argument in full: why the metric is wrong, where the money is actually going, and what an architecture-first approach would require instead.

Read Part 1 → Why NATO’s Spending Target Measures the Wrong Thing

Read Part 2 → Europe’s Defense Buildup Is Reproducing the Problem It’s Trying to Solve

This series draws on two working papers: “Institutions as Coordination Architectures” and “Market Formation as a Systems Engineering Problem.” Available on request: s@sinead.co

Subscribe now

Ireland is rich!

It ran a €42 billion budget surplus between 2022 and 2024. But, it only spends a miserly 0.2% of GDP on defense, which is the lowest rate in the European Union. In fact, it’s lower than that tiny island of Malta, and even lower than Luxembourg which is not really even a place.

Ireland’s own military has formally stated, in its own documentation, that it is “not equipped, postured, or realistically prepared to conduct a meaningful defense of the State.” Womp, womp, womp.

It probabaly seems normal then, that the political debate about this situation is overwhelmingly a debate about neutrality:

Should Ireland abandon it?
Reform the Triple Lock?
Or even… gasp!… Join NATO?

These questions dominate the airtime, the op-eds, the Dáil (parliament) debates.

And yet… the country could resolve the neutrality question tomorrow and still wake up the next morning with no radar, no air defense, naval vessels with no functioning main armament, a military at a fifty-year personnel low, a fragmented intelligence architecture unique in Europe, and a secret dependency on the Royal Air Force that nobody in public life will name out loud!

So the standard framing, that this is an ideological problem and that if Ireland just sorted out its relationship with neutrality the defense capability would follow, is wrong.

And it’s wrong in a very specific way:

My research shows that the real constraint isn’t ideology, as always, it’s that Ireland’s entire institutional system has been built around the absence of military capability.

You see, neutrality isn’t a policy choice layered on top of an otherwise normal state apparatus; it’s the premise around which the state was designed. Because eery layer of the system reflects that design:

• Political architecture (no dedicated defense ministry, no national intelligence agency),

• Legal architecture (the Triple Lock, a president who is constitutionally Supreme Commander and yet publicly opposed to the military’s existence!),

• Cultural architecture (seventy-five percent public support for neutrality as an identity commitment, not just a policy preference),

• Personnel architecture (a private’s salary that can’t compete with Dublin tech salaries),

• Procurement architecture (a system designed for routine sustainment, never for building first-generation capabilities from scratch).

This is what I unsurprisingly call an architecture problem, and it’s categorically different from a spending problem or an ideological problem.

The short answer from my research is that capability (military or otherwise) requires the co-evolution of three distinct architectures:

1. Technical (the hardware and systems),

2. Market/Industrial (the procurement pipelines, supply chains, workforce), and

3. Institutional (the political, legal, and administrative frameworks that coordinate everything else)

When these don’t co-evolve because one is missing or misconfigured, money and political will don’t convert into outcomes. Ireland has the money, and it has (albeit in fits and starts), the political will. What it lacks is the institutional architecture to convert either into military capability. And that architecture gap is, if anything, widening.

The paper applies this framework to Ireland specifically: the LOA framework the Commission on Defense Forces developed, the EU Presidency security test coming in July 2026 (spoiler: Portugal’s military gendarmerie provided drone defense during Zelensky’s visit because Ireland couldn’t), and the question of whether Ireland could actually industrialize militarily, or whether confusing engineering talent with institutional readiness is exactly the mistake the framework is designed to prevent.

Read the full paper here →

This article draws on two working papers: “Institutions as Coordination Architectures: Adaptive Bandwidth” and “Market Formation as a Systems Engineering Problem.” Both develop the formal models and cross-domain evidence summarized here. Available on request: s@sinead.co

Companion articles in this series: “Power by Other Means,” “Europe’s Defense Problem Isn’t Spending,” “How Do You Pay for Rearmament?” and “The Chips Race Is a Systems Race.”

Subscribe now

Companies have never been better equipped to understand geopolitics. Today’s multinational corporation includes Chief Geopolitical Officers, political risk subscriptions, scenario planning workshops, blah blah. The corporate intelligence industry has exploded over the past decade.

I add to this that when I moved from MIT back to Harvard, the reason was to bring geopolitical expertise into the Institute for Strategy, because the demand for this new type of intel from CEOs and boards was overwhelming.

And yet… The strategies it informs keep failing. Friend-shoring stalled; the CHIPS Act fabs got announced but ecosystems may never materialize; European hydrogen strategies produce roadmaps but not markets (I mean, it is Europe…).

So the standard post-mortem blames bad forecasting or slow execution, usually in long-winded reports that never get read, by a team that has been fired. But, my research actually shows that this common diagnosis is wrong.

The problem isn’t misreading the political environment or moving too slowly, it’s trying to execute strategies that are actually really good into markets whose supporting architectures don’t yet exist! In other words, that technical readiness (the capability to actually produce the desired thing) has outrun both market readiness and institutional readiness. The gap between strategy and outcome is structural.

I’ve spent several years at Harvard working on exactly this problem: why breakthrough technologies fail to become industries, and what the missing piece usually is.

The short answer is that market formation requires the co-evolution of three distinct architectures:

1. Technical,

2. Market, and

3. Institutional

And importantly, when one races ahead of the others, you get what I call an architecture lag, wherein the technology works, but the surrounding economic system doesn’t yet exist to support that technology.

This piece applies that framework to industrial strategy; the geoeconomic strategies companies and governments are currently betting on (friend-shoring, the CHIPS Act, European hydrogen) and asks what governments, boards, and investors should actually do differently when the problem is architectural rather than informational.

Read the full paper here →

This article applies the architecture-lag and coordination-architecture frameworks developed formally in two companion papers: “Institutions as Coordination Architectures” and “Market Formation as a Systems Engineering Problem.” Available on request: s@sinead.co

Subscribe now

Every military briefing you’ve seen in the last five years says the same thing: multi-domain operations are the future. Air, land, sea, space, cyber, electromagnetic. More is more. Integrate everything, overwhelm the enemy across all dimensions simultaneously.

And this strategy is genuinely compelling because a military force that can see, decide, and strike across six domains at once should dominate one that operates in one dimension.

But… there are tradeoffs! And what few people seem to be asking, is:

What does that actually cost?

Not the price of the platforms, because the F-35 costs what the F-35 costs.

But what I have been researching for over a decade now is: what does it cost, technology-wise and fiscally, to make all of these systems actually talk to each other in real time, under fire, across national boundaries, with fifteen-second decision windows?

Because multi-domain integration isn’t free.

If you do the math, every domain you add doesn’t just add capability, it adds complexity. And that complexity multiplies far faster than the capabilities do! Simply because of the number of interfaces that must be synchronized:

• Two domains = one interface;

• But six domains = fifteen pairwise interfaces, each with its own data links, identification protocols, timing synchronization, and rules of engagement.

And then there’s the cost inversion that very few defense economics strategists are properly accounting for:

• The integrating force pays the full coordination cost across every interface.

• But the disruptor (the adversary with a $500 drone or a GPS jammer) only needs to find one interface that fails.

Ok, so a Shahed-136 drone costs roughly $30,000. The Patriot interceptor that kills it costs $4 million. That’s a 133:1 cost ratio… and importantly….

Before you account for the coordination architecture that made the intercept possible.

This last part, the coordination architecture cost, is overwhelmingly large, and nearly always dismissed in military planning and strategy.

This is the boring stuff; the architecture that connects the radar, the command node, the launcher, and the identification system across a multinational coalition. It is also where the real expense lives.

This paper does three things:

1. Puts real numbers on the cost exchange between attackers and defenders across seven engagement types, and shows the ratios are structurally catastrophic for the integrator

2. Traces how coordination load has increased monotonically across every major coalition air campaign since Desert Storm through more threat types, more domains, less decision time, while friendly-fire incidents track coordination complexity, not technology vintage

3. Shows that only 7% of defense budgets are explicitly allocated to the coordination architecture that determines whether multi-domain integration actually works or collapses into expensive fragmentation

The punchline: multi-domain superiority is real but it is not a platform property, it is an architecture property. And architecture has a cost curve that the current defense-economic framework does not adequately measure, manage, or even see.

Read the full paper here →

The frameworks applied here — coordination cost, architecture lag, adaptive bandwidth — are developed formally in two companion papers by Sinéad O’Sullivan. Available on request: s@sinead.co

Subscribe now

In case you’ve been hiding under a rock for the last year, Europe says it has committed €800 billion to defense. And most commentary around such large financial pledges have been geopolitical:

• Who (or what) is the threat?

• What weapons will be bought?

• Is this the end of the peace dividend?

• Will this mean the end of three-month-long European summer vacations?

However, almost no one is asking the extremely exciting (I kind of joke) operational question:

How does the money actually move?

Because between a Brussels communiqué and an actually delivered fighter jet sits an entire chain of slow, shitty, boring institutional plumbing that actually “does the work” of “defense financing”:

Political authorization → Legislative instrument → Bond issuance → Treasury allocation → Procurement contract → Industrial delivery

Money can flow from capital markets to treasuries in weeks, but turning that money into signed contracts takes months to years. And as many people know, turning contracts into delivered capability takes years to decades!

It is becoming increasing important, though, for policy-makers and financiers in the defense space to understand these mechanisms for precisely the following reason:

My research around large, governmental financing programs for space and defense, and particularly last year my work on European defense, shows that the binding constraint on European rearmament is not political will, nor a lack of technologies, it is procedural capacity.

So, if we can figure out and innovate on how “defense financing” actually happens, we can finance defense better and faster!

This paper traces the full chain end-to-end, using Germany as a worked example, and breaks down:

1. Where the €800B actually comes from

2. How sovereign bond markets absorb defense issuance

3. Why borrowing costs diverge across Europe

4. How EU-level instruments like SAFE are created

5. Why procurement, not finance, is the actual source of stasis in the system

Basically, I show that until Europe builds a unified defense finance architecture that pools risk and connects capital markets directly to continental procurement pipelines, the plumbing will determine the pace, not technological progress.

Essentially, my paper that I summarize here explains how rearmament is a capital markets event, not a geopolitical one!

Read the full paper here →

The frameworks applied here, of architecture lag, premature markets, coordination architectures, are developed formally in two companion papers by Sinéad O’Sullivan. Available on request: s@sinead.co

Subscribe now

India spends roughly eight times more on defence than Pakistan (about $86 billion to Pakistan’s $10 billion). On paper, that gap should settle almost any question about relative capability, right?

But the 2025 aerial exchange over Kashmir showed something more structurally interesting; that Pakistan was able to contest the air domain credibly despite chronic fiscal distress and repeated IMF bailouts!

How? Through its industrial architecture.

More broadly, Pakistan is part of a rising class of mid-tier defence producers (à la Turkey and South Korea), whose military systems are “good enough,” entirely sovereign, and economically… cheap?

These producers are obviously not competing at the technological frontier, but where they can compete extraordinarily well is on coordination, integration, and resilience.

I’ve spent about a year getting stuck deep into mid-tier defense production and capabilities, and this research on Pakistan’s JF-17 was an outcome of the time I spent with the Pakistan Air Force last year, mapping in great detail the Pakistani industrial strategy.

In short, the 2025 India-Pakistan episode did not overturn the regional balance of power. But… it did shift the analytical lens. Military capability is increasingly an industrial property; an emergent function of how institutions, suppliers, and doctrine align over time.

Read the full paper here →

This article applies the architecture-lag and coordination-architecture frameworks developed formally in two companion papers by me, Sinéad O’Sullivan: “Institutions as Coordination Architectures” and “Market Formation as a Systems Engineering Problem.” Available on request: s@sinead.co

Subscribe now

Europe is spending more on defense than at any point in decades, which is a very good thing. In fact, EU spending reached nearly €400 billion last year, a 63% increase since Trump and Covid and the Russians scared the shit out everybody a few years ago.

Additionally, the ReArm Europe plan aims to mobilize €800 billion by 2030. It’s all happening now! Germany alone has extraordinarily managed to unlock roughly €400 billion in additional borrowing capacity for defense.

So, the will to spend, and the need to do so, is very real. And yet the system receiving this money, the EU defense industry, is structurally incapable of converting the absolutely ginormous amounts of capital into defense capabilities.

Consider that Europe fields more than five times as many variants per weapon category as the United States, and has three separate European fighter programs that have together produced roughly 1,861 aircraft.

(Don’t tell the Europeans, but the US’s F-35 program alone has generated 3,556 aircraft).

So the issue here is not spending, which we know Europe can do no problem; it’s a dangerous level of fragmentation that is currently compounding in realtime across logistics chains, spare parts, training, maintenance, and ultimately cost.

If you are a European who thinks the US DoD takes the piss when it comes to the price of an defense-rated screw, please sit down, because I’ve got bad news for you…

You know that the founding raison d’être of Europe was to have a single market? For the purpose of military and peace? Well, defense is one of the only markets where this concept breaks down entirely: the EU’s defense market is actually is comprised of twenty-seven national procurement systems, which collectively fail to generate consolidated demand. And without aggregated orders, production and manufacturing of European defense hardware remains subscale. And you guessed it: without scale, unit costs remain high. Ultimately, without competitive pricing and delivery speed, governments default outward.

So €800 billion poured into Europe’s currently fragmented system will only buy one thing: more fragmentation. Thus, the binding constraint in Europe is not capital, which we’ve (somewhat) unlocked. It is whether twenty-seven nations can construct a single institutional architecture that can convert convert €800 billion into deterrence.

Read the full paper here →

The frameworks applied here, of architecture lag, premature markets, coordination architectures, are developed formally in two companion papers by Sinéad O’Sullivan. Available on request: s@sinead.co

Subscribe now

Have you heard the latest, that space launch is cheap now?!

SpaceX fixed the space launch business, because Elon is a genius. Reusability works. The cost curve is falling. Competition will follow. Access to space is being democratized.

Except… not quite.

SpaceX’s price to launch a payload isn’t a market price, it’s a transfer price set by a vertically integrated company whose largest customer is checks notes — itself.

To rewind: In 2024, 89 of 134 Falcon launches carried Starlink; by 2025, SpaceX flew 165 missions, which is more than twice China’s entire program and roughly half of all global launches.

That cadence is insane in every possible way. But most of all because of how it impacts the economics of the launch industry, which is driven by volume! Fixed costs (by which I mean factories, pads, workforce, mission control, etc) essentially disappear at ~165 launches per year.

Which is great for SpaceX, of course.

However, fixed costs do not disappear at 7 launches a year (Ariane 6’s 2025 output).

So, who cares? Well, for the last three years, but especially in the last year, I’ve been working with governments and the private sector to figure out how, exactly, non-SpaceX’ers can become space launch companies, too. And therein lies a really important question: how can we increase domestic European or national competition in response to SpaceX? Why, for example, does Europe not have a SpaceX, too?

Cue the usual responses about Europeans being lazy, or … insert a million reasons … that essentially boil down to, because nobody has an IQ that can match Elon’s.

But here’s the BIG issue that outweighs every single other issue… In the global launch industry, competitors to SpaceX aren’t benchmarking against a market price; they’re actually benchmarking against a demand curve that they will frankly never achieve. Because SpaceX relies on induced customer demand from itself! Because the “space launch market” … doesn’t actually exist.

This is when people (including myself!) usually start muttering in low tongues about the P-word (procurement). But for a moment, consider what today’s procurement mechanisms actually do:

• Governments award on lowest cost (to SpaceX).
• Every contract to SpaceX reduces cadence for everyone else.
• Reduced cadence worsens unit economics for everyone that’s not SpaceX.
• They lose the next contract too.

The cycle is self-reinforcing, so what looks like market competition is actually the very opposite! It’s deep architectural asymmetry, meaning that long-term dependencies on SpaceX are created , which eventually, get priced accordingly.

Read the full paper here →

The frameworks applied here, of architecture lag, premature markets, coordination architectures, are developed formally in two companion papers by Sinéad O’Sullivan. Available on request: s@sinead.co

Subscribe now

Yay! Europe has found the political will to spend money on defense!

A lot, actually… €800 billion by 2030. Meanwhile NATO targets are slowly rising to 3.5% of GDP and beyond and budgets are doubling across much of the continent.

But spending is not the same as financing, which is what I discuss in this paper, the topic of which I spent much of 2025 working on in the European context.

In particular, there are specific structural issues within the EU system which mean, despite the newly found political ambition to de-couple from the US, actually driving defense spending may be harder than initially thought.

Consider the following:

• The national escape clause expires in 2028;

• SAFE is debt, not grants;

• Germany’s debt brake reform is permanent, sure, but only for Germany.

• Five member states carry debt above 100% of GDP (!)

• Defense now competes directly with dying trees and dying old people

Emergency mechanisms exist to fill the temporary spending needs, but permanent financing structures do not yet exist, meaning that in a couple of years, Europe may not be able to spend money on defense at all!

And if no durable financing architecture is built before 2028, the EU’s rearmament strategy will collides very forcefully with the same fiscal rules that constrained it for two decades. So back to square one…

Read the full paper here →

The frameworks applied here, of architecture lag, premature markets, coordination architectures, are developed formally in two companion papers by Sinéad O’Sullivan. Available on request: s@sinead.co

Subscribe now

Every few years, a battery chemistry emerges from a lab that is genuinely better than anything else on the market. Higher energy density, faster charging, longer life, safer materials, etc etc. The kind of step-change that should reshape the industry.

Yet… it almost never does?

Even when the chemistry works (or so the CEO says), the company doesn’t scale. The usual story is some combination of: the executives built the wrong factory too early, ran out of cash chasing automotive customers that take five years to say yes, got undercut by Chinese incumbents on price, or put proprietary technology in the wrong jurisdiction (read: into Asia) and lost control of it.

Once again, the problem isn’t the science or the technology, but that there is no good playbook for turning a breakthrough battery chemistry into a functioning company. And especially not in 2026, when the US policy environment, the geopolitical landscape, and the state of the industry are all moving at once and in different directions.

This paper is an attempt to write that playbook. It came out of deep strategy work I have done with exactly the battery companies navigating what I think is the most dangerous and most promising window in the battery industry in thirty years, where demand is at an all-time high; the policy tailwind is the strongest it’s ever been; and yet the industry is simultaneously falling apart.

The paper works through why the real bottleneck isn’t chemistry or capital but something much more mundane. Why the obvious first customer is the wrong first customer. And why the single most consequential strategic decision facing any US battery company right now is one that most of them are getting wrong.

Read the full paper here →

The frameworks applied here, of architecture lag, premature markets, coordination architectures, are developed formally in two companion papers by Sinéad O’Sullivan. Available on request: s@sinead.co

Subscribe now

I’ve spent the last few years watching the same thing happen over and over and over again:

A technology hits an extraordinary milestone. Like, I dunno, fusion holds plasma longer than ever, or quantum computer solves something that should have been impossible, or a lab grows a protein that didn’t exist in nature. The press coverage is breathless and a new VC fund is obviously raised in an attempt to stake a claim to this unfolding future.

But then… nothing.

The standard explanation for the lack of commercialization of these types of technologies is that they are “early.” In other words, if you give them time and let the costs come down, the private sector will eventually figure it out. And by “it”, I mean the whole commercialization thing.

But given all the time in the world, for some reason the “it” never happens. Why? Well, I suspect because the problem is not actually that the technology is early at all.

This paper introduces the core framework that I call architecture lag, which much better outlines the issues holding back commercialization across these novel technologies.

I won’t regurgitate the paper here, but I will say: if you’ve ever wondered why phenomenal technologies that clearly work keep failing to become industries, this paper outlines why. And it has important implications for how most governments currently spend their innovation budgets, as well as for how VCs intend to generate returns.

Read the full paper here →

This article draws on a companion academic paper, “Market Formation as a Systems Engineering Problem,” which develops the co-evolutionary architecture model, formal viability criteria, and cross-domain evidence in full. Available on request: s@sinead.co

Subscribe now

So here’s a question for you…

The honest answer is that it depends entirely on what you think the competition is.

If it’s a race to fabricate the most advanced chip by way of the smallest transistors, the most cutting-edge manufacturing, then the US and its allies are winning comfortably, and it isn’t close. Mostly because export controls, equipment restrictions, and the TSMC alliance have kept China roughly a generation behind at the frontier. Washington knows this, and it’s the basis of the entire chips strategy.

But here’s the thing... I keep looking at industries where one country has the best technology and another country has the best system around that technology, meaning the supply chains, the workforce, the procurement architecture, the demand coordination. And the country with the best system keeps winning.

It doesn’t happen overnight but over time, the country with the stronger surrounding system internalizes and scales the technology. By contrast, technological superiority on its own almost never overcomes systemic weakness.

This semiconductor paper applies that logic to the US–China chip rivalry. Don’t worry, I don’t re-hash the export-control debate. Instead, I ask a totally different question: what happens when the competition itself changes? Specifically, as AI shifts from frontier model training to mass deployment, what if the decisive advantages are no longer the ones the US dominates?

No spoiler alerts! drumroll… But researching this work over the space of two years reshaped how I think about what “winning” actually means in a technology contest, and left me far less certain that policymakers in Washington are framing the competition correctly.

Read the full paper here →

This article applies the architecture-lag and coordination-architecture frameworks developed formally in two companion papers by me, Sinéad O’Sullivan: “Institutions as Coordination Architectures” and “Market Formation as a Systems Engineering Problem.” Available on request: s@sinead.co

Subscribe now

Development economics has a theory of everything, it seems, except economic development.

These economists can tell you, with impressive precision, why rich countries stay rich: Secure property rights, predictable legal systems, constraints on political power… measure these, rank them, and you will get a strong predictor of long-run prosperity.

The research is influential, has won many Nobel Prizes, is widely cited, and largely correct about what it describes.

Except… It just can’t explain how countries actually get rich.

A quick clarification: when economists refer to “institutions,” they don’t mean organizations like the World Bank, the Federal Reserve, UN, etc. They mean the underlying rules and structures that govern economic life. Boring stuff like property rights, legal frameworks, regulatory agencies, procurement systems, financial regulation, civil service architecture, blah blah. In short, the apparatus through which a society coordinates economic activity.

“Good institutions,” in this sense, mean stable rules, enforceable contracts, and predictable governance.

Yet Japan, South Korea, Taiwan, Singapore (and later China) industrialized at extraordinary speed under arrangements that diverged sharply from this template! Instead of relying on free capital markets and rule-bound minimal states, they deployed state-directed credit, managed competition, bureaucratic discretion, and iterative policy experimentation. These were not not just eency teency deviations from the model; they were the growth strategy itself.

At the same time, many countries that inherited or adopted Western-style legal and administrative frameworks (e.g. parliamentary systems, codified property rights, professional civil services) failed to industrialize. The usual explanation from the economic development crowd is that these countries lacked institutional quality. I argue something different: they had institutions, but ones designed for a different problem.

This puzzle motivates the paper. The core claim is simple: we have been measuring the wrong thing!

The standard approach treats institutions as static “rules of the game” and evaluates their quality. I propose instead that institutions are better understood as coordination architectures; as systems that align investment, standards, infrastructure, finance, and governance toward structural transformation.

What separates countries that transform from those that stagnate is not institutional form, but something that I label adaptive bandwidth: the speed and flexibility with which coordination systems update when technology, markets, and governance fall out of sync.

High-bandwidth systems experiment, adjust, and reconfigure. Low-bandwidth systems, on the other hand, prioritize stability; which is valuable in mature economies, but constraining in developing ones.

This framework explains East Asian industrialization without appealing to culture or regime type, as most economists tend to do. Similarly, it explains post-colonial stagnation without resorting to institutional deficiency narratives. And it clarifies why formal convergence with Western institutional models so often fails to produce transformation.

The piece lays out the framework, the evidence, and its implications for industrial policy, productivity, and state capacity.

Read the full paper here →

This article draws on a companion academic paper, “Institutions as Coordination Architectures: Adaptive Bandwidth and the Dynamics of Economic Development,” which develops the formal model, empirical mapping, and full theoretical structure. Available on request: s@sinead.co

Subscribe now