For a debate about a country that no longer exists, this one has a surprising amount of life in it. Ask a room of socialists whether the Soviet Union was the real thing or a counterfeit, and you will not get a tired shrug. You will get an argument conducted by people who clearly still believe the answer matters.

That, I want to suggest, is something to celebrate rather than apologise for. A political tradition that keeps interrogating its own past is a confident one, not a defeated one. The very fact that the left refuses to settle this question is a sign that it holds itself to a high standard. So let us treat the debate as it deserves to be treated: as a genuinely open argument with serious people on every side.

The case that it was socialism

Start with the most straightforward reading. By the 1930s the Soviet Union had abolished the private ownership of factories, banks and large farms. There was no stock market, no class of private owners living off the labour of others, no boardroom deciding the fate of a town from a thousand miles away. The economy was planned rather than left to the market. The state described itself as a workers’ state building towards communism, and a great many sincere people around the world believed it.

The achievements were real and worth naming. A largely peasant society became literate within a generation. Industrial output that had taken Britain a century to build was assembled in decades. Women entered the workforce and the professions at a scale the West would not match for years. Healthcare and education were treated as universal entitlements rather than commodities. For millions of people who had known only the brutal hierarchies of the Tsarist empire, this felt like the future arriving early.

If socialism means the abolition of private capital and the conscious direction of an economy towards human need, the argument runs, then here it was, imperfect but unmistakable.

The case that it was state capitalism

The most influential challenge to that view came from within the socialist movement itself, which is exactly what makes it interesting. The theorist most associated with it is Tony Cliff, whose study of the question in the late 1940s asked a deceptively simple thing: a workers’ state should mean that workers actually run production, so who was running it here?

His answer was that they were not. A bureaucracy ran it. The factories were owned in name by everyone and in practice by no one the ordinary worker could hold to account. Workers could not strike, could not elect their managers, could not decide what their workplace produced or where the surplus went. The old private boss had gone, but a new collective boss had taken his place, and from the shop floor the relationship looked much the same: someone above you commanded your labour and pocketed the surplus.

On this reading, the planned economy was not free of the pressures that drive capitalism. It was locked in military and industrial competition with the West, and that competition forced relentless accumulation, investment and discipline on the workforce, just as competition between rival firms does. The bureaucracy functioned, in effect, as a single giant capitalist, and the country as a single giant firm. The forms of ownership had changed; the underlying relationship between those who command and those who labour had not.

A separate strand of this analysis grew up independently in the United States, where C.L.R. James and Raya Dunayevskaya reached a similar conclusion by a different route. That two groups arrived at the state-capitalist idea on opposite sides of the Atlantic tells you it was answering a real question, not just scoring a factional point.

The third answer: neither

There is a tidy third position that deserves a mention, because it refuses both labels. Writers such as Bruno Rizzi, and later Max Shachtman, argued that the Soviet Union was something genuinely new under the sun: not capitalism, not socialism, but a fresh kind of class society in which a bureaucratic caste, rather than private owners, sat on top. They called it bureaucratic collectivism.

Trotsky himself, watching from exile, took a halfway view. He thought the USSR remained a workers’ state because it had abolished private property, but a degenerated one, captured by a privileged officialdom that would have to be removed before the promise could be redeemed. He insisted on holding the achievement and the betrayal in the same hand, which is harder than picking a side but arguably more honest.

And honesty requires saying plainly that the betrayal was severe. The famines, the purges, the camps and the silencing of dissent are not footnotes to be managed away. Any account worth reading has to hold them. The reason the debate has such force is precisely that these things happened in a state that claimed to speak for human liberation, and people who care about that ideal have never been willing to let the claim stand unexamined.

What the argument is really about

Here is the part I find genuinely energising. On the surface this looks like a quarrel about a label for a vanished regime. Underneath, it is an argument about what socialism is for.

If you say the Soviet Union was socialism, you are accepting that socialism is compatible with one-party rule, with workers who cannot control their own workplaces, with a state that plans on their behalf but not at their command. If you say it was not socialism, you are insisting on the opposite: that socialism worthy of the name has to mean people holding real power over their own lives and their own labour, or it is not the thing at all.

So, the disagreement over a single word is really a disagreement over the goal. It forces everyone who takes part to say out loud what they actually want. And when you put it that way, you notice something hopeful: the people on both sides of this argument tend to want the same final thing. They want ordinary people to have genuine control over the conditions of their lives. They differ on whether the Soviet experiment was a flawed instance of that ideal or a departure from it, but the ideal they are measuring it against is shared.

Why the unsettled question is a strength

It would be easier, in a sense, to have a settled answer. Closed questions are comfortable. But a movement that had stopped arguing about this would be a movement that had stopped caring what its words mean, and that is a far worse place to be.

The state-capitalist tradition, whatever you make of its economics, performed a valuable service: it kept the ideal clean. It refused to let socialism be defined down to whatever a powerful state happened to be doing in its name. The socialist reading performed a different service: it insisted that the real human gains of the twentieth century not be airbrushed out by people embarrassed by the rest. Between them they hold the whole truth better than either could alone.

That is what a healthy political tradition looks like. Not one that has all the answers, but one confident enough to keep asking the hard question, to engage its sharpest internal critics in good faith, and to come out of the argument with a clearer sense of what it is really fighting for. The debate over the USSR has never been settled because the left has never stopped caring about the difference between liberation on paper and liberation in fact.

Long may it stay unsettled. A question this alive is doing exactly what a question should.

Was the USSR socialist, or state capitalist? The debate the left has never settled. was originally published in ILLUMINATION on Medium, where people are continuing the conversation by highlighting and responding to this story.

The morning of September 14, 2008, Dick Fuld was still convinced Lehman Brothers would not fail. Convinced enough that, even as Barclays walked out of the rescue talks, he kept telling colleagues the government would step in. It wouldn’t. By Monday Lehman was bankrupt, the largest in U.S. history, and a thirty-year career built on conviction had become a footnote.

Fuld was not stupid. He was not corrupt, at least not in the headline sense. He was something more dangerous than either. He was over-confident. And in that respect, he was indistinguishable from Caesar at the Rubicon, from Napoleon at Borodino, from Long-Term Capital’s Nobel laureates in 1998, from George Hudson the “Railway King,” from Theranos’s Elizabeth Holmes, from every executive who has ever signed a deck whose final slide read ”Risks: Limited.”

We tell ourselves that great collapses have many fathers — bad luck, bad timing, bad markets, bad people. They do. But sit with the case studies long enough and one variable starts to recur with the regularity of a tide. Over-confidence is not a cause of collapse. It is the catalyst. Without it, the other ingredients rarely combust.

This is uncomfortable to admit because over-confidence is also the thing most often praised in the people we promote. Conviction. Bias for action. “Disagree and commit.” Founders who refuse to listen to data. Generals who refuse to retreat. We celebrate the trait until it kills the host.

Let me show you why.

The chemistry of catalysis (not causation)

A catalyst, in chemistry, doesn’t cause a reaction. It just makes one happen faster — sometimes by a factor of millions. The other ingredients have to be present: a flammable system, a trigger, an oxygen supply. The catalyst’s job is to remove the activation barrier.

That is exactly what over-confidence does to organisations. The system was already flammable: leveraged balance sheets, brittle supply chains, ageing infrastructure, monocultures of opinion. The trigger was waiting: a default, a defection, a competitor’s product launch, a one-day liquidity crunch. The oxygen was always there: complexity that no single person can fully model.

Over-confidence removes the activation barrier. It is the reason a CEO ignores the email warning him about the leverage ratio. The reason a general ignores the scout. The reason the chief data officer waves off the model drift alert. I have been right before.

Without that catalyst, organizations slow down. They pre-mortem. They request a second opinion. They hedge. The flammable system stays flammable but does not ignite.

With that catalyst, the system always finds its match.

A short list of the catalyzed dead

- Long-Term Capital Management (1998). Two Nobel laureates and a roomful of PhDs. Their models priced volatility down even as Russia defaulted. On September 23, 1998, the New York Fed had to corral fourteen banks into a $3.625 billion rescue to keep the wreckage from spreading; the fund had lost $4.6 billion in under four months. Roger Lowenstein’s autopsy of the firm is titled, with delicious accuracy, When Genius Failed.

- Lehman Brothers (2008). Fuld personally rejected three separate rescue offers in the months leading up to the failure Korea Development Bank, Bank of America, Barclays because the terms didn’t reflect what he believed Lehman was worth. The market disagreed.

- Kodak (2012). The company invented the digital camera in 1975. It then spent thirty years protecting the film franchise because the leadership was confident, repeatedly, that digital “would never replace the quality of film.” By the time they conceded the point, the market had moved on.

- Nokia (2011–13). Internal memos from Stephen Elop’s predecessors show a leadership that genuinely believed the iPhone was a luxury toy and that Symbian’s installed base was an unassailable moat. The moat dried up in 36 months.

- WeWork (2019). Adam Neumann’s confidence was the entire product. It worked beautifully until it had to meet a prospectus.

- Caesar at the Rubicon (49 BC). The phrase alea iacta est “the die is cast” is the founding quote of over-confidence in Western history. He won. His descendants did not.

- Napoleon’s Russian campaign (1812). 685,000 men marched in. Roughly 100,000 came back. The strategic over-reach is now studied in every war college on earth as the canonical example of believing your own model.

The pattern is so consistent that you have to stop and ask: is over-confidence the common variable, or are we just seeing it because it is easy to see in hindsight? This is the right question, and it is also the question that distinguishes a curious reader from a lazy one.

None of these failures was monocausal. Strategy, incentives, timing, structure, and luck all mattered. My argument is narrower: over-confidence is often the factor that disables correction long enough for the rest to catch fire.

The counter-evidence (because this article would otherwise be propaganda)

Over-confidence is also, demonstrably, a competitive advantage in many domains.

- Startups that listen to “all the data” never get founded. The data, in year one, almost always says don’t. Survivors are people who out-believed the spreadsheet.

- Athletes who hold a realistic probability of winning Olympic gold mostly don’t win. The winners often hold an unrealistic one.

- Politicians who run national campaigns with calibrated humility lose to politicians who run them with cinematic certainty.

So, the question is not “is over-confidence bad?” that is a child’s question. The serious question is when does the same trait become a catalyst for compounding, and when does it become a catalyst for collapse?

The honest answer, after reading dozens of post-mortems, is that the difference comes down to three variables.

What separates the survivors from the catalysed

### Variable 1 — Reversibility of the bet

Jeff Bezos’s two-door framework captures this. Two-way doors (decisions you can reverse) reward over-confidence; the cost of being wrong is a step backwards. One-way doors (decisions you cannot reverse, burning ships, betting the company, invading Russia) punish over-confidence catastrophically.

The catalyzed dead almost always walked confidently through one-way doors.

### Variable 2 — Calibration over time

A founder who has been over-confident and updated every quarter is a different animal from a founder who has been over-confident and unchanged every quarter. Annie Duke calls this “resulting” judging a decision by its outcome rather than by its quality. The catalysed dead never updated. They had been right; therefore, they were right.

There is a way to measure this on yourself: write your top three predictions for the next 12 months in a notebook today. In one year, re-read. The gap between your forecasted confidence and your actual hit rate is your personal over-confidence ratio. If your confidence repeatedly exceeds your actual hit rate, you do not have conviction. You have a calibration problem.

### Variable 3 — The presence of a structural disagreer

Almost every catalyzed collapse has the same fingerprint: the disagreers were removed, ignored, or retired in the years preceding. Lehman’s chief risk officer was sidelined in 2006. Boeing’s safety engineers were reorganized under cost engineering before the 737 MAX. Theranos fired anyone who suggested the box did not work.

A leader’s over-confidence is harmless when the org has a structurally protected disagreer : a chief risk officer who cannot be fired by the CEO, an independent board chair, a “designated sceptic” at every strategy review. The moment that structural friction is removed, the catalyst meets the oxygen.

So what does the data say?

This isn’t just history dressed up as anecdote. Overconfidence is one of the most robustly documented findings in decision research. Don Moore and Paul Healy, in their much-cited paper “The Trouble with Overconfidence” (*Psychological Review*, 2008), split it into three distinct errors: overestimation (thinking you’re better than you are), overplacement (thinking you’re better than the people around you), and overprecision (being too sure your estimate is right). The third is the quiet killer. When people give a “90% confident” range for a number they don’t know, the true value falls inside it only about half the time. We don’t just guess wrong. We draw our error bars far too narrow.

It shows up in the corner office, too. Ulrike Malmendier and Geoffrey Tate, studying chief executives in the Journal of Financial Economics (2008), found that overconfident CEOs were measurably more likely to overpay for acquisitions and to destroy shareholder value doing it. The lesson isn’t “confidence is bad.” It’s that unchecked confidence concentrates risk and then doubles down on it.

Daniel Kahneman put the danger plainly: “Overconfident professionals sincerely believe they have expertise, act as experts and look like experts. You will have to struggle to remind yourself that they may be in the grip of an illusion.” The psychologist Scott Plous went further, “No problem in judgment and decision making is more prevalent and more potentially catastrophic than overconfidence.”

But here’s the part everyone skips: the cure is not less confidence. A leader frozen by self-doubt is no safer than one drunk on certainty, they just fail more quietly. The survivable mode is high confidence audibly paired with humility about the model underneath it. Philip Tetlock’s long-running forecasting work points the same way: the best predictors hold “strong opinions, weakly held,” and they update fast when reality pushes back.

The catalyzed dead held strong opinions, strongly held.

A working checklist before your next big call

Whether you are a fund manager, a CEO, a head of engineering, or a graduate student about to defend a dissertation, here is a five-question screen I now run before any decision I cannot easily reverse.

1. What would the smartest, friendliest sceptic say is wrong with this?
If I cannot generate that view in 60 seconds, I haven’t done the work.

2. Who in this organisation is structurally allowed to tell me I’m wrong, and have they been heard this week? If the answer is “no one” or “no,” the catalyst is loose.

3. Is this a one-way door or a two-way door? If one-way, divide my confidence by two and ask the question again.

4. What is the dollar / time / reputation cost of being wrong, and am I provably willing to pay it? Vague willingness is not willingness.

5. What is my personal over-confidence ratio over the last 12 months? If I haven’t measured it, my conviction is fictional.

This list will not stop you from being over-confident. Nothing will. But it will let you spot the catalyst in your own bloodstream before it meets the oxygen.

The uncomfortable conclusion

We have a cultural ritual of blaming collapses on bad people, bad markets, or bad luck. The post-mortems sell better that way. They were greedy. They were unlucky. They were villains.

The truth is duller and more frightening. Most of them were people very much like us competent, hard-working, well-spoken, well-credentialed operating in flammable systems, surrounded by removable disagreers, marching confidently through one-way doors. The catalyst was already in them.

It is in us too. The question is not whether you carry it (you do), but whether you have built the structural friction that prevents it from finding its match.

Over-confidence is the catalyst. The other ingredients are always around. The work, the only work that consistently matters at scale is to keep the activation barrier intact.

Caesar didn’t think the river was that wide. Fuld didn’t think Sunday would come. The next person on this list is reading something exactly like this article right now and quietly disagreeing.

I hope it isn’t you. I’m not sure it isn’t me.

Is Over-Confidence the Catalyst for Every Great Collapse? was originally published in ILLUMINATION on Medium, where people are continuing the conversation by highlighting and responding to this story.

There has never been a better time to start something.

You can build a website without being a professional developer. You can design a product without hiring a designer. You can write a book without waiting for a publisher. You can even build software with tools that would have seemed impossible a few years ago.

Artificial intelligence has lowered the barrier to entry for almost everything.

And yet, look around.

People have hundreds of unfinished projects. Folders full of business ideas. Bookmarks containing tutorials they will never finish watching. Subscriptions to tools they barely use. And notes filled with plans for a future that never seems to arrive.

We have become incredibly good at preparing to succeed.

But preparation and progress are not the same thing.

The new form of procrastination looks productive

Procrastination used to be easy to recognize.

You watched television instead of working. You played games instead of studying. You spent the afternoon doing absolutely nothing.

Now procrastination can look like work.

You spend three hours comparing AI writing tools. You ask an AI to generate twenty business ideas. You research the most profitable niche, study your competitors, redesign your logo, and create an elaborate plan for your launch.

At the end of the day, you feel exhausted.

But what have you actually built?

Sometimes the answer is nothing that another person can use, buy, experience, or benefit from.

This is the strange trap of the AI era: you can experience the feeling of productivity without producing anything meaningful.

Every new tool offers the possibility of a better result. Every new tutorial suggests that you are one missing piece of knowledge away from success.

So you keep searching.

Not necessarily because you are lazy, but because searching feels safer than exposing your work to the world.

An idea cannot be rejected. An unfinished product cannot receive a bad review. A business that never launches cannot prove that your assumptions were wrong.

Potential is comfortable.

Reality is not.

AI made the beginning cheap. The ending is still expensive.

Consider two people who want to create a digital product.

The first person spends a month researching profitable niches, testing prompts, comparing platforms, and asking AI to improve the business plan.

The second person chooses one specific problem, creates a small solution, publishes it, and shows it to potential customers.

The first person may know more about digital products.

The second person has something the market can respond to.

That difference matters.

The second person can discover whether customers understand the offer, whether the price makes sense, whether the product solves a real problem, and whether people are willing to pay.

The first person is still operating on assumptions.

This does not mean research is useless. It means research has a point of diminishing returns.

At some stage, additional information stops reducing uncertainty as effectively as taking action does.

You cannot think your way into knowing how the market will respond to a product. Eventually, you have to put the product in front of the market.

AI can help you get there faster.

It cannot remove the need to get there.

The uncomfortable truth about having access to powerful tools

Imagine giving two people access to the same AI system.

Both can generate business plans, write persuasive copy, produce images, create websites, and analyze information.

One person creates something useful and improves it through feedback.

The other creates increasingly sophisticated plans for things they have not launched.

After six months, the difference between them probably will not be explained by who had access to the better prompt.

It will be explained by what they did with the results.

Tools create possibilities. Habits determine which possibilities become real.

This is why I am becoming less interested in asking, “Which AI tool is the best?”

A more useful question is:

Which tool helps me finish something that matters?

The distinction sounds small, but it changes how you choose what to learn.

You stop collecting tools because they look impressive. You start choosing tools because they remove a specific obstacle between you and a finished result.

You stop asking AI to generate endless possibilities. You ask it to help you evaluate one possibility, build it, test it, and improve it.

You stop measuring progress by the number of things you know how to do.

You begin measuring it by the number of useful things you have completed.

Completion is a competitive advantage

Finishing sounds ordinary. It is not as exciting as discovering a revolutionary strategy or finding an overlooked business opportunity.

But completion creates something that endless preparation cannot: evidence.

A finished article teaches you what readers respond to.

A published product teaches you whether people understand its value.

A working website reveals problems that a design mockup cannot.

A real conversation with a potential customer challenges assumptions that hours of market research may leave untouched.

Each completed project gives you information you can use in the next one.

Over time, this creates a compounding advantage.

You become better at estimating how long work takes. You learn which mistakes you repeatedly make. You discover where your standards are too low and where perfectionism wastes your time. You develop judgment that cannot be acquired entirely through tutorials.

Most importantly, you stop treating every new project as a completely unfamiliar experience.

You have been there before.

You know what it takes to move from an idea to a result.

That knowledge is valuable precisely because it comes from experience rather than imagination.

Stop building your identity around what you could do

There is another reason people struggle to finish things.

An unfinished project protects the identity they want to believe in.

“I could build a successful business.”

“I could write a great book.”

“I could become a skilled developer.”

“I could create something people love.”

All those statements remain possible while the person avoids testing them.

But once you finish and publish something, you encounter reality. The work may be mediocre. Nobody may notice it. Your first attempt may fail to generate a single sale.

That experience can be painful because it separates the person you imagine yourself to be from the person your current skills and habits allow you to be.

The solution is not to avoid that gap.

The solution is to use it.

You cannot improve an imaginary skill through imaginary practice. You improve by producing work, observing the result, and correcting what went wrong.

Your first version does not need to prove that you are exceptional.

It needs to give you something real to improve.

And if the result is disappointing, that is information not a permanent verdict on your potential.

A practical experiment: finish one small thing

If you recognize yourself in this article, do not respond by creating another complicated productivity system.

Try a smaller experiment.

Choose one project you have been postponing. It could be an article, a portfolio website, a digital product, a useful tool, or a small business offer.

Then follow four rules.

1. Define what “finished” means.

Make the outcome specific enough that you can tell whether it exists.

“Work on my business” is vague.

“Publish one product page with a clear offer and a working purchase link” is concrete.

2. Remove unnecessary decisions.

Choose your tools, set a reasonable scope, and stop reopening decisions that do not materially affect the result.

Not every choice deserves another hour of research.

3. Use AI to overcome obstacles, not create more options.

Ask for help with the problem directly in front of you. If you are writing, get feedback on the draft. If you are coding, debug the error. If you are creating a product, improve the offer.

Do not turn every obstacle into an excuse to redesign the entire project.

4. Put the result in front of someone.

Publish it, share it, test it, or ask a potential customer to evaluate it.

Completion does not guarantee success. It gives you the opportunity to learn something that matters.

Then use that information to decide what comes next.

You do not need to finish everything you start. Some ideas deserve to be abandoned. But you should learn to distinguish a project you have deliberately rejected from one you have simply avoided finishing.

Those are two very different things.

The next advantage will not belong to everyone with AI

AI will continue to improve. Tasks that currently require specialized knowledge will become easier. More people will be able to create things that previously required teams, money, or years of training.

That is genuinely exciting.

It also means that access to the tools themselves may become less distinctive.

When thousands of people can generate a similar website, article, design, or business plan, the ability to generate it is only part of the equation.

Knowing what is worth creating matters.

Understanding the person you are creating it for matters.

Recognizing bad output matters.

Building trust matters.

And being willing to take a project across the finish line through the tedious revisions, uncomfortable feedback, and imperfect first release matters enormously.

The opportunity is not simply to become better at using AI.

It is to become better at turning what AI makes possible into something another human being actually values.

That requires judgment, persistence, and contact with the real world.

No prompt can substitute for all three.

One final thought

You probably do not need another hundred ideas.

You may not need another tool, another course, or another strategy.

You might need to take one thing you already know how to do and carry it far enough that it becomes real.

Not perfect. Not guaranteed to succeed. Not necessarily impressive to everyone watching.

Just finished, useful, and available for the world to respond to.

Because in a world where starting is becoming almost effortless, the person who can consistently finish meaningful work may have an advantage that no new tool can hand them.

The question is no longer just what you could create. It is what you are willing to finish.

The Internet Doesn’t Need Another AI Expert. It Needs People Who Can Finish Things. was originally published in ILLUMINATION on Medium, where people are continuing the conversation by highlighting and responding to this story.