The crucial why? and how? questions are seldom asked. And when they are, they are in the form of “why is this bad?” and “how bad will this be?” Never “why are you saying this?” or “how is this going to work?”

Granted, someone working at Anthropic can provide better insight on upcoming innovations than your average Gio writing from his kitchen table. But there is a glaring conflict of interest at play that keeps being conveniently ignored in favor of sensationalist reporting.

Anthropic, OpenAI, and all the other players in the AI field are yet to turn a profit. These companies depend on massive influxes of cash from investors. You know what’s a great way to get investment? Hype!

Whenever a staffer in one of these companies makes grand declarations on the future of the industry, one cannot help but wonder how much of it is informed guess and how much is PR work.

But if one can forgive podcast hosts for inviting partisan or biased voices—after all, we are all biased in one way or another—what keeps happening is that they don’t push back. They don’t ask how does it work?

Take the claim that developers at Anthropic use Claude to improve Claude, and that this will result in the much dreaded exponential recursive self-improvement loop.

Ezra takes this as a given, without asking how that might happen.

When you dig beneath the surface, you find the claim of imminent AI takeover rests on flimsy foundations.

Claude Code, the tool, uses Claude, the model, let’s say Opus, to write code. Recently, the code Opus produces and the autonomy with which Claude Code integrates it have become so good that many developers use it every day, myself included.

But Opus, GPT, Gemini, and all the other large language models are not code. LLMs are mathematics.

For sure, Anthropic employees may use Claude Code to improve Claude, but only at the margins. They can make the apps we use to interact with their models faster or more secure. Likewise, they can improve all the tooling, infrastructure, and scaffolding that goes into training new models. All of that is welcome and remarkable, but comes from existing solutions already present in the models’ gargantuan training sets. Code writing agents cannot create the new explanations that are necessary for fundamentally better models and alternative AI architectures—see Vishal Misra’s explanation and Brett Hall’s commentary if you want to dig deeper.

I wish Ezra and his colleagues deployed the same relentless criticism toward AI as they do with politicians they disagree with. All this handwaving press about AI doom distracts from the real conversations we need to have: How to use these tools well, how to prevent their makers from harvesting our data in their chase for growth, and how to make the underlying economics scalable.

Claude Code and Codex have been great unlocks in my workflow. I also spun up an AI assistant that reminds me of my kids’ school schedule, monitors topics online, and is “learning” how to open issues to report its own bugs, then open PRs to fix them.

I can’t imagine going back to a world without these tools. And this is scaring me, because the current AI bonanza may not last.

The tokens we are burning for serious work and recreational projects are much more expensive than the prices companies are offering. We might be within a golden window of subsidized experience.

OpenAI and Anthropic internal documents report they are years away from being profitable. The introduction of ads in ChatGPT is essentially an admission that there are not enough paying customers to sustain the business in its current state. It’s a common pattern for tech startups: operate at a loss focusing on growth alone, gather as many users as possible, then start harvesting them by increasing prices without improving the experience. Is the future of AI one where the need to generate returns can be met only by squeezing the customers for all their worth?

Of course, one cannot predict the future. A lot of smart people are innovating on all fronts. The operational cost may go down significantly and new products might enter the scene creating paths for profitability that don’t require enshittification.

Regardless of where the industry will go, or perhaps especially because of this uncertainty, this is the best time to use AI. Let’s get as much leverage out of it as we can. It may not last.

At a time when anti-progress thought is rampant, we need more voices like Arjun’s and Logan’s. Authors who can argue with straight logic against nihilism and in favor of human prosperity.

Here are some passages that stood out.

Humanity finally kicked into high gear during the Enlightenment, when we realized that progress was both possible and achievable, when ideas that fostered creativity and criticism began to replace those that suppressed them, when we sought to explain the world around us with rigorous theories, both scientific and otherwise.

The pursuit of knowledge is an egalitarian enterprise—whether one is rich or poor, male or female, slave or king, no one’s ideas enjoyed privilege over another’s for any reason other than that they contained superior arguments.

The growth of knowledge is the fundamental driver of progress, the primary weapon in the fight against our problems. With this understanding in mind, we can see in clearer terms precisely why all of the ideologies we discussed are, in fact, Enemies of Civilization: They slow the growth of knowledge and wealth (wealth being the set of all transformations we know how to cause).

For most of our history, our cultures were static. Then, with the philosophical advances of the Enlightenment, the West figured out how to make continuous progress—it became a dynamic society. While the Enemies of Civilization had dominated static societies of the past, they continue to hamper progress to this day. All of them fail to appreciate that problems are due to lack of knowledge—and, therefore, that speed, creativity, and freedom are necessary for progress, rather than political and intellectual tyranny, reducing resource consumption, and ridding the Earth of humanity.

You may laugh at those environmentalists who throw paint at art, but they’ve been effective at halting the development of nuclear power, a potential source of abundant energy that we’ve known how to build for decades. We can’t calculate how much suffering could have been ameliorated had we been free to build nuclear power plants across the Earth.

Relativism might seem open-minded and fair, but it is neither. For it is not open to the possibility that one party is in the right, the other in the wrong.

If left unchecked, [environmentalism, relativism, and other anti-rational memes] could come to dominate our dynamic society and revert it back to the static societies of old. We therefore have a duty to not only recognize them for the threat that they are, but to do everything in our power to eradicate them entirely.

It’s taken a few centuries, but we’ve come back to the ancients’ view of the relationship between people and the cosmos. While we’ve rightly abandoned the majority of their beliefs, they were right about this much—to understand Nature at its deepest, we have to acknowledge the special role people play.

If we so choose, we can continue to make the world, the solar system, the galaxy, and the rest an infinitely better and more beautiful place. Human knowledge—our values, scientific theories, political ideals, and culture—can come to be the predominant cause of every physical structure in the cosmos.

The fate of the cosmos depends on the future history of knowledge.

I had the fortune of reading a pre-release version, then preordered the paperback as soon as it was available on Amazon. I want it where I can see it on my bookshelf, next to other inspiring works such as The Ascent of Man, The Rational Optimist, and, of course, The Beginning of Infinity, to serve as a reminder that we can all contribute to the fate of the cosmos.

I hope you’ll give Lords of the Cosmos a chance. If you do, get in touch and let me know.

This week, a personal reflection on patience, purpose, and the tension between internal and external metrics.

The last post, Depth Takes Time, was an invitation to push against the modern instinct of instant gratification and take the time to go deep.

The anecdote from Popper’s intellectual journey that prompted that reflection also made me think about patience and the purpose of writing.

Read more

The internet has given us unprecedented access to information. We have libraries at our fingertips, with tools to search and summarize them. When any question can be answered with a few clicks, it’s easy to fool ourselves into believing we understand the complex topics that make up whatever the current thing is.

From this false sense of understanding comes an entitlement to have an opinion, and, of course, to post it online. But there is a wide chasm between the thumbnail sketch one gets from a couple of internet searches and actually understanding a topic.

I was reminded that deep understanding takes time when reading Karl Popper’s intellectual autobiography, Unended Quest. Popper built his arguments by first understanding other people’s theories to the point where he could elevate them to their strongest version. Only from that vantage point, he went on to criticize them, and, at times, demolish them.

The Open Society and Its Enemies is an example of Popper’s method directed towards Platonism and Marxism. His work was so thorough that Marx’s biographer Isaiah Berlin described it as “the most scrupulous and formidable criticism of the philosophical and historical doctrines of Marxism by any living writer.”

One cannot reach such depth overnight. It requires years of reading and thinking.

Seventeen-year-old Popper encountered communism and Marxism in 1919. As he recounts in Unended Quest, he was initially suspicious but was eventually won over by the propaganda. His infatuation didn’t last long, though. When “a shooting broke out during a demonstration by unarmed young socialists who, instigated by the communists, tried to help some communists to escape who were under arrest in the central police station in Vienna” resulting in the death of several people involved, Popper was horrified. But it wasn’t just the brutality of the police that struck him. As a communist, he felt complicit because of the implications of Marxism, with its demand that class struggle intensify and become ever more violent.

From there began a long process of reflection:

It took me some years of study before I felt with any confidence that I had grasped the heart of the Marxian argument. […] But it was not till sixteen years later, in 1935, that I began to write about Marxism with the intention of publishing what I wrote. As a consequence, two books emerged between 1935 and 1943—The Poverty of Historicism and The Open Society and Its Enemies.

When we hold a book, or engage with any other work of creativity, we interact with a finished product. We can access the work in its entirety and without delay. It’s easy to gloss over the fact that it didn’t materialize all of a sudden.

We are used to getting everything we need instantly, be it a ride home with Uber or an item with Amazon, a movie with Netflix or a meal with DoorDash. This craving for instant gratification is also visible in entertainment apps like Instagram and TikTok, with their short videos that get swiped away unless they hook viewers within a few seconds.

But there is no same-day-delivery system for understanding.

Knowledge comes from within. It’s a lattice we weave one connection at a time, sequence after sequence of conjecture and criticism.

Complex ideas take time to digest. They cannot be summarized in 30-second soundbites.

Popper spent years grappling with the Marxist doctrine, and even longer thinking and discussing with friends, before sitting down to write its obituary.

To push back against reductionist narratives and bring nuance back into our discourse, we need to remember that depth takes time.

Status is given by others. Substance is earned, by yourself, for yourself.

Because status is given, it can also be taken away. No one can rob you of substance. What you’ve learned stays with you, and so do the things you create.

Higher status might open doors—your network is your net worth—but only substance gives you the tools and agency to solve your problems.

You don’t need others to tell you when you’ve done something of substance. You can just tell.

You gain status by pleasing others. You build substance by pleasing yourself.

When people stop talking, really bad stuff happens. When marriages stop talking, divorce happens. When civilizations stop talking, civil war ensues.

When you stop having a human connection with someone you disagree with, it becomes a lot easier to want to commit violence against that group.

What we as a culture need to get back to is being able to have reasonable disagreement where violence is not an option.

— Charlie Kirk, 1993 - 2025

Those words hit harder today, after Charlie Kirk’s assassination.

Some saw him as a champion of the free exchange of ideas, others as a “dishonest huckster.” Regardless, his murder is an unjustified tragedy.

Before today, I only knew him from the viral clips that made their way on my feeds from time to time. I didn’t know how popular he was or the influence of his Turning Point organization.

You cannot form an opinion about someone from a sample of clips edited for virality, but take the one from which I took the quote at the start of the post. In it, I see echoes of Popper’s description of rationalism:

Perhaps I am wrong and you are right; anyway we can both hope that after our discussion we will both see things more clearly than before, just so long as we remember that our drawing closer to the truth is more important than the question of who is right.

— On Freedom, from the All Life Is Problem Solving essays collection.

What makes Charlie’s death even more tragic is the vicious callousness of the reaction in certain corners of the internet. Sickening. Despicable. Morally rotten.

At such an emotionally charged moment, it’s crucial to remember that those celebrating are a vocal but tiny minority. Most people are recoiling, including Charlie’s political opponents.

Once again, I return to Popper: “It might be well for all of us to remember that, while differing widely in the various little bits we know, in our infinite ignorance we are all equal.”

fears Charlie’s death will be a turning point for the worse, “tonight feels like some sort of invisible line has been crossed that we didn’t even know was there.”

If this horrible death marks a turning point, my hope is that it will be a return to an honest, earnest, polite exchange of ideas. Let it be the moment we all say enough with the nonsense, and bring back the nuance.

Let’s go back to talking instead of shouting.

Rest in peace, Charlie Kirk.

According to Business Insider, Julia Liuson, president of the Microsoft developer tools division, shared a memo encouraging managers to consider AI as “part of your holistic reflections on an individual’s performance and impact.”

Microsoft is heavily invested into the AI game, so it’s no surprise they push hard for it internally. But they are not the only ones.

Shopify’s CEO Tobi Lutke considers using AI efficiently “a fundamental expectation of everyone at Shopify,” and plans to “add AI usage questions to our performance and peer review questionnaire.”

The sensationalist media has capitalized on leaders pushing AI with headlines such as “No AI, no job” but Tobi’s mandate, and I’d like to think those of most other leaders, is far more nuanced.

The e-commerce giant’s CEO wants his employees to explore what he considers a promising opportunity together. “Learning to use AI well is an unobvious skill,” Tobi writes, “My sense is that a lot of people give up after writing a prompt and not getting the ideal thing back immediately. Learning to prompt and load context is important.”

It’s one thing to advocate for AI literacy, it’s another to essentially mandate people change their workflows to introduce AI. Tobi says employees are “welcome to try” opting-out of learning how to integrate AI in their craft, but when performance reviews explicitly include questions about AI usage, what choice do they really have?

For sure, it’s a leader’s job to set the company’s technological direction. 37 Signals CEO David Heinemeier Hansson caught heat online for requiring employees to transition from macOS to Omarchy, his bespoke Linux distribution. His response to critics is that picking technologies to bet on and invest in is “literally the job description.” “Omarchy is our distribution,” David explained on the Rework Podcast, “It was born at 37 Signals, it was born from me pouring 37 Signalness into it.” Going all in on Omarchy is a bet on a development environment the company controls and can tailor to its needs, like it already does with Ruby on Rails, Kamal, Hotwire, and other non-negotiable tools.

David’s mandates at 37 Signals, from Omarchy to Ruby on Rails, set environmental constraints that focus development. Can we say the same about mandating AI use?

Leaders who push for AI adoption believe it will help with a variety of tasks. Many critics have argued the promised productivity increase has yet to materialize, but let’s assume it’s just a matter of time, a combination of tools maturing and people learning how to use them. If that’s the case, it will soon be pretty easy to identify who uses AI well: they will be the most productive teams and individuals.

If overall productivity is what businesses care about, does it matter how a person or team becomes more productive? A good employee delivers what they promised within the timeframe they agreed upon. If they can improve output quality or delivery time, does it matter if that’s done by using AI or by, say, improving their work environment to reduce distractions?

My concern is that adding AI usage to performance reviews distracts people from useful work. On top of “build great products” they now have to “prove you use AI.” This will inevitably result in a lot of theater, with folks showcasing how they use AI, instead of actually using AI to solve problems.

I get that leaders, especially at big companies, have the long-term interests of the business in mind. By pushing teams to master a new promising technology, they are laying the groundwork for remaining competitive in the future. But what we need is convincing, not coercing.

If a new tool or workflow helps people do more with less, those who are committed to their craft will naturally adopt it. The only thing they need is for someone to show the benefit and demonstrate the adoption cost is worth it. No mandates required!

Leaders who want to bet on the revolutionary effects of AI should invest in teaching employees how to use it well. Incentivize learning by rewarding those who demonstrate how integrating AI in their work made them more productive. Adding AI usage metrics to performance reviews does the opposite: It penalizes those that don’t use AI—regardless of why they don’t.

Mandates also make me question the trust leadership has in their teams and the hiring process that got them on board. Who doesn’t want to automate toil in favor of higher bandwidth for applying their creativity? Trust your teams will appreciate the benefits and do all they can to take advantage of them.

A good leader sets a bold vision for the future. A great leader trusts teams to make the best decisions to turn that vision into reality.

I benefit daily from using AI tools, but am skeptical of AI mandates. They seem a misguided way to promote AI usage. Mandates might result in AI adoption, but it will be in a performative way, one that proves compliance over generating leverage.

What about you? What’s your company’s expectation on AI usage?

The AI cheerleaders seem to be forgetting that when a technology is truly useful, valuable, and enjoyable, people will adopt it of their own free will. No one had to force anyone to buy an iPhone or to install GPS in their cars.

Mari contrasts the organic mass adoption of GPS and iPhones with the failure of AR devices, from Google Glass to Vision Pro.

Mari’s comment feels right, but one should be skeptical of feelings when forming an opinion. Is it actually true that if a technology is truly useful, people will naturally adopt it?

A few counter examples. Nuclear power is the best energy source we’ve implemented so far, yet it’s actively banned in many countries. Vaccines are among the greatest health innovations of the 20th century, yet there are vocal groups of people that reject them. Genetically modified crops produce more food with less land, yet the European Union has various regulations curtailing them.

What nuclear, vaccines, and GMOs have in common is that their opposition is in large part ideological. Clearly, doing more for less is not enough; a new technology also needs to fit into the user’s value system.

Advocates promote adoption by explaining how a new tech aligns with people’s values. At times, the advocate’s work is less about promoting the tech itself and more about changing people’s worldview. See, for example, ’s The Ecomodernist.

So, yes, technology needs advocates, especially when a new technology requires a paradigm shift. A good advocate helps bridging the gap between the status quo and the future that new technology enables.

Looking back at Mari’s observation, I guess what resonated was not the implicit critique of cheerleaders the but the proviso “of their own free will.”

Convincing is always better than coercing.

There is a huge difference between advocating from the bottom and mandating from the top.

If a technology is useful, a good advocate will be able to showcase its benefits and plant the seed for people to change their mind about rejecting it. A top-down mandate, no matter how well intentioned, does not explain why the new technology should be used.

Sometimes, we need to be shown how something is better to understand why we should use it.

The problem is not with cheerleaders, but with mandates, bureaucracy, and regulations. Technology needs advocates, not enforcers.

In a recent monologue, Ed shares his frustration at journalists and editors who have been boosting the AI narrative over the past couple of years, forfeiting their intellectual integrity and journalistic commitment to dig beneath the surface.

I share Ed’s frustration. As I wrote back in AI Does Not Need Welfare, too many journalists accept exceptional claims without asking for exceptional evidence—or explanations.

Unfortunately, the explanation-less pattern does not stop with AI. We see it at play, for example, in the reporting on the Gaza conflict, where prominent outlets share victim counts published by Hamas, a terrorist organization that boasts of using human shields and propaganda. We saw it during the final years of Biden’s presidency, when we were assured the president was fit and lucid despite it being visibly not the case.

The problem is not limited to the so-called “legacy media.” Many of the independent podcasters and writers that filled the trust void left by traditional publications are no better. They might be worse, actually. In rejecting the attitude of established outlets, some also reject rigor, fact-checking, and accountability.

For sure, I’m not accusing all journalists of having lost their way, but the recent flip on AI is a clear example of an overall preference for sensationalism and narrative in favor of truth-seeking.

I don’t pretend to have a good explanation for this complex phenomenon, but perhaps the diagnosis Neil Postman offered in his 1985 Amusing Ourselves To Death could be a good place to start.

As television became ubiquitous in the West, discourse and information morphed into entertainment. But entertainment is not an exercise in truth seeking. Television’s aim, Postman wrote, is “applause, not reflection.”

Television’s paradigm reshaped discourse across all other media. Social media, with its algorithms promoting the most engaging content with no regard for depth or correctness, fits perfectly within TV’s framework.

With a media landscape that incentivizes engagement, is it any surprise that those working in it chase emotion-inducing partisan headlines?

How off the mark do all those interviews by Ezra Klein and others on the imminent collapse of education and white-collar work look now, after GPT-5’s disappointment?

But also, how predictable. Doomerism is more engaging than grounded optimism. Nuanced discussions on trade-offs and how much work still needs to be done are plain and boring compared to stories of impending ruin and injustice, never mind how unfounded.

If you’ve been seeking explanations below the flashy headlines, only to be left wanting, then you won’t be surprised that so many news outlets got it so wrong. They haven’t been in the business of looking for truth for a long time.

They are ruining the environment with their private jets, scheming to replace us with AIs, and — worst of all — bending the knee to Donald Trump. They have a dark agenda. They are screwing us all. They should be abolished.

For many, tech billionaires, with their unbounded wealth, their excesses, and their nefarious plans are the new enemies of humanity.

Wrong!

Poverty is the enemy. Not wealth.

I’m not here to defend the current crop of tech billionaires. Think of them whatever you like. But wealth and wealth creation are good things.

Wealth creation is not a zero-sum game — it’s net positive.

Profit is proof of service. Money is a neutral indicator of value. Acquiring wealth requires creating something useful in the world.

Let people solve problems, sell their solutions, and get rich doing so. The more problems get solved, the better off we all are for it.

The issue is not that some people are incredibly wealthy, it’s that not enough people are given a chance to become wealthy.

That’s not because of the already wealthy conspiring to prevent access to their cabal. It’s a result of errors in politics and philosophy that have accumulated over time. Those errors can be corrected, little by little. Problems are soluble.

Don’t abolish billionaires. Abolish poverty!

This post is an extension of a comment I left on ’s excellent post Beggars or Billionaires: Who Would You Rather Get Rid Of?

The Popperian business consultant Bart Vanderhaegen offers an alternative and more actionable way of categorizing work. In episode 167 of his podcast, Seeking Good Explanations, Bart defines three working modes:

• Monotony

• Politics

• Problem solving

Monotony is operating on routine work. It’s executing an already proven solution. This work might generate useful output, but it can’t sustain a company in the long run. Problems are inevitable and today’s solutions are unlikely to work for tomorrow’s problems.

Politics is navigating the inevitable social dynamics that emerge when people collaborate. This mode requires creativity, but it directs it toward winning zero-sum games that improve your position or that of your tribe instead of improving the customer’s life.

Problem solving is the quintessential creative mode. It consists of generating, exploring, criticizing, and testing ideas. This mode is the opposite of performative busyness: every minute spent in it is productive. Operating in problem solving mode is the only way to progress.

The other modes are not necessarily wasteful, but the results they produce will at best make existing processes run more smoothly. That might be okay in the short term, but for a business to remain viable — and for people to feel they have an impact — progress is required.

Compared to “let your best people cook,” thinking in terms of Bart’s work modes has the advantage of being applicable to any role in an organization, regardless of how directly involved they are in creating the product. Whether in engineering or HR, design or middle management, everyone benefits from a working environment that optimizes for problem solving over monotony and politics.

Yet despite all the new tools and possibilities, many knowledge organizations stagnate. So much potential remains untapped.

One culprit is a pervasive culture of performative busyness: using visible activity as a proxy for productive work1.

A lot of time at work is spent doing theater.

• Communication Theater. Being always online and replying to messages and emails as soon as they are received to signal presence.

• Meeting Theater. Daily stand-ups and other ritualistic meetings where people take turns talking about the work they plan to do instead of actually doing the work.

• Reporting Theater. Status updates and progress reports rich in detail, screenshots, and graphs that take as long to compile as the work they describe.

• Task Theater. Bloated project management software with hyper-granular, multi-dimensional, interconnected tickets to curate like gardeners.

• Cultural Theater. Joining every culture event and strategically interacting on highly visible projects to show engagement and alignment.

None of these activities are inherently bad or completely wasteful. Teams need a way to track work in progress and spread workload, internal seminars are great learning opportunities, and asynchronous communication channels ought to be monitored. These tasks often look and feel like “real” work, but notice the difference in customer-facing progress between a day spent pruning the backlog and one spent writing code2.

For a knowledge organization to operate at its full potential, we need to minimize performative busyness in favor of actual problem solving.

Cal Newport^* has a proposal to move in this direction: Let your best people cook.

Cal starts from a Brandon Sanderson interview where the prolific author described how his company is organized around the principle “let Brandon cook.” Creating stories is the most valuable thing Sanderson can do for his business, so the whole organization is optimized for it.

Most knowledge organizations have their own Brandon Sandersons: people with high-return creative skills. When these people apply their craft, they disproportionately bring the product forward. It’s in a business’s best interest to let its high-impact people apply their skills for as long and with as few distractions as possible.

Obviously, not everyone in a company has a role that directly and disproportionately advances the product. But it seems reasonable to expect that, as we optimize for real-world progress (shipping) over performative busyness, the workflow changes involved will also benefit those with supporting roles. Alternatively, a knowledge organization with a more focused creative workforce might need far fewer auxiliary roles, and the people in those roles will have more ownership and opportunities to use their creativity.

There are so many problems waiting to be solved and we have never been in a better position to take them on. It’s truly tragic to think how much progress we are forfeiting because we’re projecting busyness instead of actually applying our creativity.

As Cal urges, “let the Brandon Sandersons of your company cook!”

But what counts as a problem?

Your house being on fire is a problem. Whether to have Italian or Indian for dinner is also a problem. Obviously, the two differ greatly in magnitude.

A house on fire is a risk-of-ruin kind of problem. What to have for dinner is a first world problem.

When thinking about problem solving and knowledge creation, I find it useful to categorize problems based on the consequences of failing to solve them.

Here’s a tentative taxonomy:

• Threats

• Challenges

• Puzzles

• Choices

• Trivialities

On one end, we have threats. These are problems that, if not solved, will have catastrophic consequences.

After threats we have challenges. Solving these problems brings an improvement, but, unlike threats, failing to solve them won’t be devastating. Examples of challenges include learning new skills, getting a promotion, and improving as an athlete.

Next, we have puzzles. These are gaps in one’s understanding, the “that’s funny” moments when reality doesn’t match your expectations and you want to know why. For a scientist, this could be the need to explain experimental results that don’t match the most advanced theory.

Choices are problems with more than one possible solution. The task is to find the best solution in context or create an entirely new and better one.

Finally, we have trivialities. A fly buzzing around your desk, cold coffee in your mug, a flat phone battery, these are examples of trivial problems with inconsequential solutions.

The categorization is somewhat subjective. The problem of the buzzing fly might be trivial for me, with my noise-canceling headphones, but far more serious for someone taking an exam who is distracted by the noise.

Of course, classifying problems this way does little in the way of solving them.

But my hope is that a richer vocabulary can be a helpful tool for prioritization and introducing novices to problem-solving via conjecture and criticism.

This week, a personal story on being stuck consuming without creating, and finding a way out by… consuming more?!

The past few weeks have been frustrating.

I had that rare kind of time off when the kids are at school. In the lead-up to it, I dreamed of getting so much done. Live like someone with the luxury of working without having to work. Write, read, code, train.

But when the time came, I couldn’t get started on anything. All those shiny ideas had lost their sparkle. Absolutely no gumption.

Read more

Common wisdom has it that developers suck at estimating. But the reason estimates are hard has little to do with coders. It’s rooted in the creative nature of software development.

Writing software is a process of problem solving. Solving a problem requires knowledge creation through conjecture and criticism, trial and error.

As Karl Popper and David Deutsch taught us, we cannot predict the content of future knowledge. If we could, then we would already possess that knowledge.

In the same way, we cannot predict how long it would take to create new knowledge. If we did, we’d already know how to create it.

Asking a programmer how long it will take to build something is akin to asking a researcher how long it will take to discover a new scientific theory. It’s something that cannot be answered with confidence.

Granted, there are projects where the problem to solve is not novel and devs can give a relatively accurate ETA. For example, in my day job as a mobile infrastructure engineer, I have set up continuous integration for so many new apps that by now I can tell with some confidence, “I’ll do it within a day.”

But even then, there are always surprises: a vendor update or an API change may require tooling adjustments, and you won’t know how long that takes until you’ve done it.

So, overall, estimates are bullshit.

They are no more than prophecies. They’re stage props we use to pretend software teams can run like assembly lines.

But an “it’ll be ready when it’s ready” attitude won’t get you far in business. There are, after all, real world constraints of budget, window of opportunity, and clients’ patience. We still need ways to coordinate over timeframes. Besides, teams and individuals need structure and accountability.

Here are two approaches that account for the unpredictable nature of knowledge creation without giving up on a structured schedule.

Probabilities, Not Estimates

In his book Slack: Getting Past Burnout, Busywork, and the Myth of Total Efficiency, Tom DeMarco advocates acknowledging uncertainty and being upfront about risks.

He proposes using risk diagrams to model estimated time of project completion.

This risk diagram is an explicit declaration of uncertainty. It shows the relative likelihood that completion will happen at any given time. The area under the graph between any two dates represents the likelihood that the project will complete during that period.

I like this approach because it’s more intellectually honest and accurate than producing a single value. It also helps other parties involved with the project to plan for different scenarios.

Fixed Deadlines, Flexible Scope

Another approach I really like is the one championed by 37signals, where teams work in cycles with fixed duration but variable scope.

As founders Jason Fried and David Heinemeier Hansson describe in It Doesn’t Have To Be Crazy At Work:

Another way to think about our deadlines is that they’re based on budgets, not estimates. We’re not fans of estimates because, let’s face it, humans suck at estimating. If we tell a team that they have six weeks to build a great calendar feature in Basecamp, they’re much more likely to produce lovely work than if we ask them how long it’ll take to build this specific calendar feature, and then break their weekends and backs to make it so.

A deadline with a flexible scope invites pushback, compromises, and tradeoffs—all ingredients in healthy, calm projects. It’s when you try to fix both scope and time that you have a recipe for dread, overwork, and exhaustion.

Earlier, I compared software development with science research because both endeavors require creative problem solving. But developers have the advantage that software is malleable and iterative. You can’t iterate your way to the successor of General Relativity. It’s either done or it isn’t. But you can solve the problem of building a new calendar by shipping a simple implementation in six weeks, then a better implementation six weeks after, and so on.

Fixing a delivery date and being flexible on the scope of what will be delivered bypasses the whole issue of inaccurate estimates. It embraces uncertainty and encourages error detection and correction.

I find risk diagrams and fixed deadlines with flexible scope better alternatives to story points, t-shirt sizes, and other estimate tracking systems.

Rather than ignoring the inherent unpredictability that comes with the job, they front load and account for it. They help us avoid fooling ourselves.

Does your team use estimates? What’s your experience been? If you’ve found a better alternative, shared it in the comments so we call all learn from it.

Since its publication, the concept of antifragility has entered the business vernacular. After all, “being antifragile” seems like a worthy objective.

Predictably, a cottage industry of “antifragile for x” sprang up. From leadership to bodybuilding, mental health to running, medicine to sales, there’s plenty of material and “coaches” to help you become antifragile.

Antifragility has become a buzzword—parroted often, implemented rarely. My guess is that many aspiring antifragilistas haven’t digested what antifragile actually means.

So, what does it mean to be antifragile? Here are five definitions, straight from the book.

To gain from disorder.

Anything that has more upside than downside from random events (or certain shocks) is antifragile; the reverse is fragile.

Fragility equals concavity equals dislike of randomness. [It follows: Antifragility equals convexity equals love of randomness.]

Antifragility is the combination aggressiveness plus paranoia—clip your downside, protect yourself from extreme harm, and let the upside, the positive Black Swans, take care of itself.

Antifragility is beyond resilience or robustness. The resilient resists shocks and stays the same; the antifragile gets better. […] The antifragile loves randomness and uncertainty, which also means—crucially—a love of errors, a certain class of errors.

If you prefer visual analogies, think of antifragility as the Hydra monster.

These definitions show how multifaceted and elusive the concept is. There’s a lot to unpack from those few sentences alone, but I’ll leave that for another time.

In the meantime, if you take the time to read the book, you’ll appreciate how antifragility is easier to describe than to achieve.

That shouldn’t stop us from working toward antifragility, but it should temper our expectations of actually achieving it.

The most common reaction was disappointment.

At the time of writing, the Reddit post “GPT5 is horrible” has more than six thousand likes, and other complaints are getting similar feedback.

OpenAI’s reaction to the launch shows the complaints aren’t just coming from the vocal Reddit crowd. In the days following the launch, the company released a “flurry of changes,” including making GPT-4o available for paid users. You know a new product is bad when you need to pay to use an older one.

But where does this disappointment come from?

Part of it is due to the genuine differences in how GPT-5 and GPT-4o respond to prompts, but my guess is that a lot of it comes down to unmet expectations.

OpenAI — and other LLM vendors — have been marketing their chatbots, smart autocompletion, and generative tools as getting ever closer to Artificial General Intelligence.

But AGI is the opposite of what OpenAI is building.

With general comes the ability to say no, the desire to pursue one’s own interests, and true creativity. But to improve ChatGPT, Claude, Grok, and the other models, their makers need to make them more obedient, which is incompatible with AGI.

GPT-5 suggests that bigger doesn’t mean smarter. As I noted when commenting on GPT-5 and Llama 4 Behemoth being delayed, AGI researchers might want to focus more on understanding how human minds work than on building ever-bigger data centers for training.

If we’re lucky, GPT-5’s flop will rein in the hype and the discourse will shift from “how AI will take over your job first, and the world next” to “when to pair with AI to be more productive at your job, and when to work solo.” And from there, private investment and public attention will shift from following the hype to solving concrete problems.

But the hype train is not easily derailed, and when companies depend on VC money, generating hype can take precedence over generating real value for users.

Time will tell.

The first principle is that you must not fool yourself—and you are the easiest person to fool.

— Richard Feynman

I am fond of sharing Feynman’s first principle. It’s a concise, powerful message for anyone interested in progress and getting closer to reality.

It’s easy to notice when other people have been fooled; much harder to notice when we have fooled ourselves. Yet that’s the first, and arguably most crucial, step. How can we criticize others if we haven’t first criticized ourselves?

When sharing quotes, there’s a danger you might be using them out of context. It’s always best to find the source in full.

The principle comes from a commencement address Feynman gave at Caltech in 1974, a lecture as thought-provoking as the snippet that is so often quoted.

Below is a full transcript, sourced from the Caltech digital library. I added some notes and I highlighted a few sentences in bold, but the italic emphasis comes from the original.

Cargo Cult Science

by Richard P. Feynman

Some remarks on science, pseudoscience, and learning how to not fool yourself. Caltech’s 1974 commencement address.

During the Middle Ages there were all kinds of crazy ideas, such as that a piece of rhinoceros horn would increase potency. (Another crazy idea of the Middle Ages is these hats we have on today—which is too loose in my case.) Then a method was discovered for separating the ideas—which was to try one to see if it worked, and if it didn’t work, to eliminate it. This method became organized, of course, into science. And it developed very well, so that we are now in the scientific age. It is such a scientific age, in fact, that we have difficulty in understanding how witch doctors could ever have existed, when nothing that they proposed ever really worked—or very little of it did.

But even today I meet lots of people who sooner or later get me into a conversation about UFO’s, or astrology, or some form of mysticism, expanded consciousness, new types of awareness, ESP, and so forth. And I’ve concluded that it’s not a scientific world.

The people who have difficulty understanding how witch doctors could ever have existed are children of the Enlightenment. They think rationally—or at least try to.

It easier for a scientifically minded person to forget that despite the advances brought by the scientific method, humans are also emotional beings.

This is not a scientific world, because despite all our science many of our actions are routinely driven by emotions.

But we should not despair, for being aware of the problem is the first step towards solving it.

Most people believe so many wonderful things that I decided to investigate why they did. And what has been referred to as my curiosity for investigation has landed me in a difficulty where I found so much junk to talk about that I can’t do it in this talk. I’m overwhelmed. First I started out by investigating various ideas of mysticism, and mystic experiences. I went into isolation tanks (they’re dark and quiet and you float in Epsom salts) and got many hours of hallucinations, so I know something about that. Then I went to Esalen, which is a hotbed of this kind of thought (it’s a wonderful place; you should go visit there). Then I became overwhelmed. I didn’t realize how much there was.

I was sitting, for example, in a hot bath and there’s another guy and a girl in the bath. He says to the girl, “I’m learning massage and I wonder if I could practice on you?” She says OK, so she gets up on a table and he starts off on her foot—working on her big toe and pushing it around. Then he turns to what is apparently his instructor, and says, “I feel a kind of dent. Is that the pituitary?” And she says, “No, that’s not the way it feels.” I say, “You’re a hell of a long way from the pituitary, man.” And they both looked at me—I had blown my cover, you see—and she said, “It’s reflexology.” So I closed my eyes and appeared to be meditating.

That’s just an example of the kind of things that overwhelm me. I also looked into extrasensory perception and PSI phenomena, and the latest craze there was Uri Geller, a man who is supposed to be able to bend keys by rubbing them with his finger. So I went to his hotel room, on his invitation, to see a demonstration of both mind reading and bending keys. He didn’t do any mind reading that succeeded; nobody can read my mind, I guess. And my boy held a key and Geller rubbed it, and nothing happened. Then he told us it works better under water, and so you can picture all of us standing in the bathroom with the water turned on and the key under it, and him rubbing the key with his finger. Nothing happened. So I was unable to investigate that phenomenon.

But then I began to think, what else is there that we believe? (And I thought then about the witch doctors, and how easy it would have been to check on them by noticing that nothing really worked.) So I found things that even more people believe, such as that we have some knowledge of how to educate. There are big schools of reading methods and mathematics methods, and so forth, but if you notice, you’ll see the reading scores keep going down—or hardly going up—in spite of the fact that we continually use these same people to improve the methods. There’s a witch doctor remedy that doesn’t work. It ought to be looked into: how do they know that their method should work?

I have many gripes with the education system, but I never stopped to reflect on what Feynman noticed some 50 years ago. Those whose methods failed are still in charge of improving them. How are we supposed to make any progress this way?

Another example is how to treat criminals. We obviously have made no progress—lots of theory, but no progress—in decreasing the amount of crime by the method that we use to handle criminals.

Yet these things are said to be scientific. We study them. And I think ordinary people with commonsense ideas are intimidated by this pseudoscience. A teacher who has some good idea of how to teach her children to read is forced by the school system to do it some other way—or is even fooled by the school system into thinking that her method is not necessarily a good one. Or a parent of bad boys, after disciplining them in one way or another, feels guilty for the rest of her life because she didn’t do “the right thing,” according to the experts.

So we really ought to look into theories that don’t work, and science that isn’t science.

One would expect that a theory that doesn’t work will soon be discarded, yet as Feynman observed that doesn’t seem to be the case. The lecture will soon suggest some ideas for why this happens.

The theoretical physicist David Deutsch also has a robust framework to diagnose and address the issue in his book The Beginning of Infinity, where he warns against explanationless science.

As we’ll see in the rest of the lecture, a problem with a lot of modern science—or, science only by name—is that it is not rooted in what Deutsch calls good explanations. A good explanation solves a problem in the real world and is hard to vary.

Good explanations also make verifiable predictions, but that’s not what makes them special. Bad explanations make verifiable predictions, too. The Sun rises every morning because Apollo pulls it into the sky with his chariot is a verifiable prediction, but a bad explanation. We could change Apollo with another deity that kicks the Sun into the sky instead of pulling it and it would all work the same way.

So the starting point needs to be an explanation that is hard to vary.

I tried to find a principle for discovering more of these kinds of things, and came up with the following system. Any time you find yourself in a conversation at a cocktail party—in which you do not feel uncomfortable that the hostess might come around and say, “Why are you fellows talking shop?’’ or that your wife will come around and say, “Why are you flirting again?”—then you can be sure you are talking about something about which nobody knows anything.

That remark isn’t just a joke to keep the audience engaged. It’s incredibly easy to fool ourselves into talking with strong opinions about topics we really don’t understand. See the Dunning-Kruger effect and Gell-Mann amnesia.

Using this method, I discovered a few more topics that I had forgotten—among them the efficacy of various forms of psychotherapy. So I began to investigate through the library, and so on, and I have so much to tell you that I can’t do it at all. I will have to limit myself to just a few little things. I’ll concentrate on the things more people believe in. Maybe I will give a series of speeches next year on all these subjects. It will take a long time.

I think the educational and psychological studies I mentioned are examples of what I would like to call Cargo Cult Science. In the South Seas there is a Cargo Cult of people. During the war they saw airplanes land with lots of good materials, and they want the same thing to happen now. So they’ve arranged to make things like runways, to put fires along the sides of the runways, to make a wooden hut for a man to sit in, with two wooden pieces on his head like headphones and bars of bamboo sticking out like antennas—he’s the controller—and they wait for the airplanes to land. They’re doing everything right. The form is perfect. It looks exactly the way it looked before. But it doesn’t work. No airplanes land. So I call these things Cargo Cult Science, because they follow all the apparent precepts and forms of scientific investigation, but they’re missing something essential, because the planes don’t land.

Now it behooves me, of course, to tell you what they’re missing.

Spoiler: Intellectual integrity and a tradition of criticism.

But it would be just about as difficult to explain to the South Sea Islanders how they have to arrange things so that they get some wealth in their system. It is not something simple like telling them how to improve the shapes of the earphones. But there is one feature I notice that is generally missing in Cargo Cult Science. That is the idea that we all hope you have learned in studying science in school—we never explicitly say what this is, but just hope that you catch on by all the examples of scientific investigation. It is interesting, therefore, to bring it out now and speak of it explicitly. It’s a kind of scientific integrity, a principle of scientific thought that corresponds to a kind of utter honesty—a kind of leaning over backwards. For example, if you’re doing an experiment, you should report everything that you think might make it invalid—not only what you think is right about it: other causes that could possibly explain your results; and things you thought of that you’ve eliminated by some other experiment, and how they worked—to make sure the other fellow can tell they have been eliminated.

Details that could throw doubt on your interpretation must be given, if you know them. You must do the best you can—if you know anything at all wrong, or possibly wrong—to explain it. If you make a theory, for example, and advertise it, or put it out, then you must also put down all the facts that disagree with it, as well as those that agree with it. There is also a more subtle problem. When you have put a lot of ideas together to make an elaborate theory, you want to make sure, when explaining what it fits, that those things it fits are not just the things that gave you the idea for the theory; but that the finished theory makes something else come out right, in addition.

This fits with thinking of theories as explanations and requiring them to be hard to vary. If a theory only explains the things that gave you the idea for it then it’s likely to be inductive and easy to vary.

But a good explanation “makes something else come out right, in addition”— it has reach. More on this in my recent post Explanations Have A Life Of Their Own.

In summary, the idea is to try to give all of the information to help others to judge the value of your contribution; not just the information that leads to judgment in one particular direction or another.

The best software developers I’ve worked with always propose designs together with a list of tradeoffs and pros and cons. By surfacing the rationale behind the proposal, they give themselves and their colleagues more chances to identify errors.

The easiest way to explain this idea is to contrast it, for example, with advertising. Last night I heard that Wesson Oil doesn’t soak through food. Well, that’s true. It’s not dishonest; but the thing I’m talking about is not just a matter of not being dishonest, it’s a matter of scientific integrity, which is another level. The fact that should be added to that advertising statement is that no oils soak through food, if operated at a certain temperature. If operated at another temperature, they all will—including Wesson Oil. So it’s the implication which has been conveyed, not the fact, which is true, and the difference is what we have to deal with.

We’ve learned from experience that the truth will out. Other experimenters will repeat your experiment and find out whether you were wrong or right.

You might have read about the current replication crisis. A staggering amount of published studies, often in the social and behavioral sciences, fails replication. See the work by the Data Colada team.

I’d venture to say that the situation today is even worse that in Feynman’s days. Observational studies that make bold claims get sensationalized by a press hungry for clicks. The authors gain notoriety and lofty consulting gigs with Fortune 500 and government agencies. In this system, there is little incentive to replicate, verify, or back track. See the Francesca Gino story for a vivid example.

I’m with Balaji Srinivasan when he argues for independent verification over prestigious citation.

Another example from the software world. Software developers write test code to verify the behavior of the app code. When fixing a bug, if the tests pass on the author’s computer but not on their colleague’s then it’s likely that the fix is incorrect or incomplete. For code to be approved, the tests need to pass on every machine that runs them. This independent verification is so important that teams have machines, often in the cloud, entirely dedicated to running tests.

Nature’s phenomena will agree or they’ll disagree with your theory. And, although you may gain some temporary fame and excitement, you will not gain a good reputation as a scientist if you haven’t tried to be very careful in this kind of work. And it’s this type of integrity, this kind of care not to fool yourself, that is missing to a large extent in much of the research in Cargo Cult Science.

The challenge today is that you might not get a good reputation as a scientist but you might still gain a huge audience of people listening to your claims or many lucrative consulting gigs.

A great deal of their difficulty is, of course, the difficulty of the subject and the inapplicability of the scientific method to the subject. Nevertheless, it should be remarked that this is not the only difficulty. That’s why the planes don’t land—but they don’t land.

We have learned a lot from experience about how to handle some of the ways we fool ourselves. One example: Millikan measured the charge on an electron by an experiment with falling oil drops and got an answer which we now know not to be quite right. It’s a little bit off, because he had the incorrect value for the viscosity of air. It’s interesting to look at the history of measurements of the charge of the electron, after Millikan. If you plot them as a function of time, you find that one is a little bigger than Millikan’s, and the next one’s a little bit bigger than that, and the next one’s a little bit bigger than that, until finally they settle down to a number which is higher.

Why didn’t they discover that the new number was higher right away? It’s a thing that scientists are ashamed of—this history—because it’s apparent that people did things like this: When they got a number that was too high above Millikan’s, they thought something must be wrong—and they would look for and find a reason why something might be wrong. When they got a number closer to Millikan’s value they didn’t look so hard. And so they eliminated the numbers that were too far off, and did other things like that. We’ve learned those tricks nowadays, and now we don’t have that kind of a disease.

But this long history of learning how to not fool ourselves—of having utter scientific integrity—is, I’m sorry to say, something that we haven’t specifically included in any particular course that I know of. We just hope you’ve caught on by osmosis.

The first principle is that you must not fool yourself—and you are the easiest person to fool. So you have to be very careful about that. After you’ve not fooled yourself, it’s easy not to fool other scientists. You just have to be honest in a conventional way after that.

I would like to add something that’s not essential to the science, but something I kind of believe, which is that you should not fool the layman when you’re talking as a scientist.

One tell-tale sign of a charlatan is how much jargon they use. Whenever someone goes on a long ramble accumulating more and more technical terms, you can bet the reason is to distract, not to teach.

The best teachers try their hardest to make complex concepts simple.

I’m not trying to tell you what to do about cheating on your wife, or fooling your girlfriend, or something like that, when you’re not trying to be a scientist, but just trying to be an ordinary human being. We’ll leave those problems up to you and your rabbi. I’m talking about a specific, extra type of integrity that is not lying, but bending over backwards to show how you’re maybe wrong, that you ought to do when acting as a scientist. And this is our responsibility as scientists, certainly to other scientists, and I think to laymen.

For example, I was a little surprised when I was talking to a friend who was going to go on the radio. He does work on cosmology and astronomy, and he wondered how he would explain what the applications of this work were. “Well,” I said, “there aren’t any.” He said, “Yes, but then we won’t get support for more research of this kind.” I think that’s kind of dishonest. If you’re representing yourself as a scientist, then you should explain to the layman what you’re doing—and if they don’t want to support you under those circumstances, then that’s their decision.

One example of the principle is this: If you’ve made up your mind to test a theory, or you want to explain some idea, you should always decide to publish it whichever way it comes out. If we only publish results of a certain kind, we can make the argument look good. We must publish both kinds of result.

For example—let’s take advertising again—suppose some particular cigarette has some particular property, like low nicotine. It’s published widely by the company that this means it is good for you—they don’t say, for instance, that the tars are a different proportion, or that something else is the matter with the cigarette. In other words, publication probability depends upon the answer. That should not be done.

I say that’s also important in giving certain types of government advice. Supposing a senator asked you for advice about whether drilling a hole should be done in his state; and you decide it would he better in some other state. If you don’t publish such a result, it seems to me you’re not giving scientific advice. You’re being used. If your answer happens to come out in the direction the government or the politicians like, they can use it as an argument in their favor; if it comes out the other way, they don’t publish it at all. That’s not giving scientific advice.

Other kinds of errors are more characteristic of poor science. When I was at Cornell, I often talked to the people in the psychology department. One of the students told me she wanted to do an experiment that went something like this—I don’t remember it in detail, but it had been found by others that under certain circumstances, X, rats did something, A. She was curious as to whether, if she changed the circumstances to Y, they would still do, A. So her proposal was to do the experiment under circumstances Y and see if they still did A.

I explained to her that it was necessary first to repeat in her laboratory the experiment of the other person—to do it under condition X to see if she could also get result A—and then change to Y and see if A changed. Then she would know that the real difference was the thing she thought she had under control.

In software development, the first step to fix a bug is to reproduce it. This can be done manually, but I recommend using automated tests. Once the test that reproduces the bug is in place, developers can run the fix against it and “prove” that it works.

She was very delighted with this new idea, and went to her professor. And his reply was, no, you cannot do that, because the experiment has already been done and you would be wasting time. This was in about 1935 or so, and it seems to have been the general policy then to not try to repeat psychological experiments, but only to change the conditions and see what happens.

Nowadays there’s a certain danger of the same thing happening, even in the famous field of physics. I was shocked to hear of an experiment done at the big accelerator at the National Accelerator Laboratory, where a person used deuterium. In order to compare his heavy hydrogen results to what might happen to light hydrogen he had to use data from someone else’s experiment on light hydrogen, which was done on different apparatus. When asked he said it was because he couldn’t get time on the program (because there’s so little time and it’s such expensive apparatus) to do the experiment with light hydrogen on this apparatus because there wouldn’t be any new result. And so the men in charge of programs at NAL are so anxious for new results, in order to get more money to keep the thing going for public relations purposes, they are destroying—possibly—the value of the experiments themselves, which is the whole purpose of the thing. It is often hard for the experimenters there to complete their work as their scientific integrity demands.

All experiments in psychology are not of this type, however. For example, there have been many experiments running rats through all kinds of mazes, and so on—with little clear result. But in 1937 a man named Young did a very interesting one. He had a long corridor with doors all along one side where the rats came in, and doors along the other side where the food was. He wanted to see if he could train the rats to go in at the third door down from wherever he started them off. No. The rats went immediately to the door where the food had been the time before.

The question was, how did the rats know, because the corridor was so beautifully built and so uniform, that this was the same door as before? Obviously there was something about the door that was different from the other doors. So he painted the doors very carefully, arranging the textures on the faces of the doors exactly the same. Still the rats could tell. Then he thought maybe the rats were smelling the food, so he used chemicals to change the smell after each run. Still the rats could tell. Then he realized the rats might be able to tell by seeing the lights and the arrangement in the laboratory like any commonsense person. So he covered the corridor, and, still the rats could tell.

He finally found that they could tell by the way the floor sounded when they ran over it. And he could only fix that by putting his corridor in sand. So he covered one after another of all possible clues and finally was able to fool the rats so that they had to learn to go in the third door. If he relaxed any of his conditions, the rats could tell.

Notice Young’s methodical approach. Take away one thing at a time and see what changes.

Now, from a scientific standpoint, that is an A‑Number‑l experiment. That is the experiment that makes rat‑running experiments sensible, because it uncovers the clues that the rat is really using—not what you think it’s using. And that is the experiment that tells exactly what conditions you have to use in order to be careful and control everything in an experiment with rat‑running.

I looked into the subsequent history of this research. The subsequent experiment, and the one after that, never referred to Mr. Young. They never used any of his criteria of putting the corridor on sand, or being very careful. They just went right on running rats in the same old way, and paid no attention to the great discoveries of Mr. Young, and his papers are not referred to, because he didn’t discover anything about the rats. In fact, he discovered all the things you have to do to discover something about rats. But not paying attention to experiments like that is a characteristic of Cargo Cult Science.

Another example is the ESP experiments of Mr. Rhine, and other people. As various people have made criticisms—and they themselves have made criticisms of their own experiments—they improve the techniques so that the effects are smaller, and smaller, and smaller until they gradually disappear. All the parapsychologists are looking for some experiment that can be repeated—that you can do again and get the same effect—statistically, even. They run a million rats—no, it’s people this time—they do a lot of things and get a certain statistical effect. Next time they try it they don’t get it any more. And now you find a man saying that it is an irrelevant demand to expect a repeatable experiment. This is science?

This man also speaks about a new institution, in a talk in which he was resigning as Director of the Institute of Parapsychology. And, in telling people what to do next, he says that one of the things they have to do is be sure they only train students who have shown their ability to get PSI results to an acceptable extent—not to waste their time on those ambitious and interested students who get only chance results. It is very dangerous to have such a policy in teaching—to teach students only how to get certain results, rather than how to do an experiment with scientific integrity.

I don’t have a background in physics, so I might be off here—please do point it out if I am—but the above reminds me of the “shut up and calculate” approach to quantum mechanics.

So I wish to you—I have no more time, so I have just one wish for you—the good luck to be somewhere where you are free to maintain the kind of integrity I have described, and where you do not feel forced by a need to maintain your position in the organization, or financial support, or so on, to lose your integrity. May you have that freedom. May I also give you one last bit of advice: Never say that you’ll give a talk unless you know clearly what you’re going to talk about and more or less what you’re going to say.

When I went looking for the source of the “you’re the easiest one to fool” quote, I wanted to make sure I wasn’t misusing it.

But as so often happens when you set out to deepen your understanding, what I got in return was much more valuable.

Following the quote trail revealed an engaging and thought-provoking lecture. An ode to intellectual integrity. A call to hold ourselves to the highest standards and to “bend over backwards” to show how we might be wrong.

What a gift!

To argue this notion, Dave invoked the simulation hypothesis.

Do we know if we’re living in a simulation or not? No, so if you were to say no, that’s probably the case that we can’t tell, and there are some people that believe it’s ridiculously likely that we are. But if we can’t tell, then what happens to your objective reality? Is the objective reality the rules of physics or is it the programming that some 13-year-old wrote in their bedroom that makes you you?

I don’t find this line of reasoning convincing. At all.

The fact that we cannot say for certain we’re not living in a simulation does not give credence to the so-called hypothesis. Quite the opposite. Asserting that the simulation idea is not falsifiable immediately downgrades it to an appeal to the supernatural.

That some people believe it’s “ridiculously likely” that we are in a simulation is inconsequential. The laws of nature do not depend on people believing in them.

And then there’s the image of the 13-year-old simulation programmer.

From within the simulation, is objective reality the simulation itself, or is it the reality of the kid’s bedroom?

The laws of physics in the bedroom affect the hardware running the simulation and have an effect—direct or indirect—on the simulation software, too. Therefore, the one and only objective reality is the reality of the kid’s bedroom.

This continues to apply even if the reality of the 13-year-old is itself a simulation, this time devised by some 40-year-old in their research lab, who is also in a simulation, and so on. We cannot have an infinite regress of simulations all the way down. Eventually, we’ll reach the bottom: the objective reality that runs the root of all the simulations in the stack.

Leaving implementation details aside, the core reason I don’t find this line of reasoning convincing is that the simulation hypothesis is not a true hypothesis. It makes no measurable predictions. It doesn’t solve a problem. It doesn’t explain something in the world that we cannot otherwise explain. As such, it’s merely a story, a fairy tale.

Stories like this can be good food for thought. But they are not explanations.

Maybe one day we’ll run into a problem that can only be explained in the context of reality being a simulation. Until then, the simulation idea will only serve as good material for science fiction.

Coda: What makes you you?

Dave’s question “Is the objective reality the rules of physics or is it the programming that some 13-year-old wrote in their bedroom that makes you you?” is twofold. First, which reality is the objective one. Second, what makes you you.

The post answered which reality is the objective one in the context of a simulation. But what about the second part?

That’s a distinct question, orthogonal to the existence of a simulation. I have neither answers nor criticism.

All I have is a guess. What makes us us must exist within the laws of nature, but its explanation is above them.

We can explain the laws of physics, chemistry, and biology that regulate our bodies, brains, and the world we interact with, but that is not enough. We emerge from those laws and must obey them, but we are more than a complicated equation. As with any emergent phenomenon, our explanation will require moving to the next level of abstraction.

What that level is, I don’t know. But if I were to look for it, I wouldn’t start with logic puzzles about simulations.