For ten minutes, a Cambridge master’s student sat down with Claude to study Plato’s Allegory of the Cave.

I then handed him a quiz to test his understanding. He aced the recall questions. On the application questions, he struggled. Between the study session and the post-test debrief, his confidence score had dropped two points.

He was the third of four people I watched do almost exactly the same thing.

If you’ve used Claude or ChatGPT to study something recently, you’ve probably had a version of this experience where the chatbot’s explanation feels crystal clear to you in the moment, you nod along, and you close the tab feeling smarter. Only to then sit down to recall or apply the knowledge shortly after and you find yourself underprepared.

It’s not you. It’s that AI is exceptional at making you feel like you’ve learned something but not as great at making sure you actually have. To build durable understanding, you still have to get active in your own learning.

I arrived at this conclusion from a class I took at Stanford last fall called ‘Technology for Learners’ taught by Professor Candace Thille, who’s been building the design infrastructure behind edtech for over two decades.

She founded the Open Learning Initiative at Carnegie Mellon and later at Stanford, co-authored the U.S. National Education Technology Plan, and even left academia for a stint at Amazon as Director of Learning Science and Engineering, where she was responsible for how 1.6 million employees learned on the job.

The capstone for her class was an ALT (Analysis of a Learning Tool) where we picked a tool, ran a study, and asked whether it helped people actually learn.

I chose an AI tool because, at Stanford, AI is choking us out. It writes emails, automates our lives, matches us with dates, and (even though our honor code forbids it) still helps people with assignments and, yes, exams.

To the point that the business school has brought back handwritten exams for some courses. Yup, it’s that bad.

But can we really blame students for using AI when their teachers are using it too, as Anthropic’s recent study shows?

As a Claude ambassador on campus, I needed to know if this thing I’m promoting is actually helping us learn or just liquefying our brains. Also because Claude was still lesser known on campus compared to Gemini and ChatGPT at the time, it was an opportunity to get more people to try it out.

So for my ALT, I focused on a specific question: does Claude support durable learning, the kind that sticks with you when you need to use it to solve problems?

I observed 4 people (the Cambridge graduate student, a Stanford law student, and two young professionals in big tech and venture capital), each over a 10-minute study session on Plato’s Allegory of the Cave.

Then I gave them a test mapped to Bloom’s Taxonomy, covering everything from basic recall to applying the knowledge in new contexts. Afterwards, we debriefed.

I chose a philosophical text because I felt it’d have the right balance of abstraction and familiarity for a group of motivated young learners to feel fairly challenged.

Here’s what I found.

The Illusion of Competence

Before AI, learning had natural friction built in.

You had to find the textbook, parse the dense prose, re-read the confusing paragraph, put it down, come back, explain it to a friend, get it wrong, and try again.

This is what cognitive scientists call “desirable difficulties,” the productive struggles that force our brains to actually encode information rather than just passively absorb it. As my friend Lenore writes, if you want to learn durably, you can’t afford to skip the reps.

But Claude removes that struggle almost entirely by being extremely good at explaining things clearly and going no further. But for durable learning, some friction is necessary.

Back to the Cambridge student.

He was the most prepared participant in my study by a wide margin. A long-time Claude power user with a custom communication style for his Claude pre-configured to challenge him while he learned. He came into the study session ready to ace the test, but not the learning.

On recall, he scored the highest. He could name the core elements of the allegory, identify what the prisoners represented, and describe what the journey outside symbolized. But when asked to construct his own allegory using Plato’s framework, pressed for time, he struggled. By the time we got to the debrief, his self-rated confidence in the material had dropped two points from where it was right after his study session.

The problem was that what he’d actually learned was Claude’s clean, coherent, well-organized explanation and not his own original mental model of the allegory itself.

Three of the four participants observed the same pattern of a clear mismatch between how confident they felt after the 10-minute session with Claude and how much they’d actually learned after they reflected on the learning session post-test.

Claude created what I called concept fluency fallacy, the feeling that the material is clear, coherent, and familiar even when your underlying mental model—the real substrate of learning—is still not yet fully formed.

For me, questions come from noticing a gap between a topic and my mental model of it, then making a conscious effort to close that gap. This is why I see asking questions, the foundation of Socratic inquiry, as the hallmark of active learning.

The Expertise Paradox

Another thing I found in my research is that learning with AI is far easier (and far more effective) if you already know a lot about the thing you’re trying to learn.

Weird, right?

But that’s because the quality of what you get from Claude is almost entirely a function of the quality of the questions you ask; and the quality of your questions is almost entirely a function of how much you already know.

Also, true experts on a topic are less likely to mistake concept fluency for real learning because they have a working mental model to measure against.

If learning is what happens when the unknown becomes known, then the fastest way to learn is to use what you already know as scaffolding for what you don’t, using analogies to link unfamiliar concepts onto familiar contexts and mental models.

The tech professional in the study showed me this in real time.

During his session, he noticed Claude making connections between Plato’s allegory and feminist epistemology. He got curious. If Western feminist theory could help explain Plato, then African philosophy should be able to as well.

He then asked Claude to explain the allegory through the lens of Ubuntu, the African concept of interconnectedness. Claude ran with it and engaged him in a rich, cross-cultural dialogue that let him use his prior mental model as scaffolding to understand the concept more deeply.

To no one’s surprise, he had the highest nominal and confidence scores after being tested on the material.

What This Means for How We Learn

Because my study had only four participants, this is far from exhaustive.

But the patterns I noticed were consistent enough to take seriously, and they point toward a product design problem that makes learning with AI less durable than it could be.

Claude, like every other AI chatbot, currently conflates the two fundamentally different jobs in a single interface: problem-solving and learning.

The uniformity of the experience subconsciously encourages shallow interaction patterns, which is fine when you need a fast answer, but not when you’re trying to build deep, durable understanding on a topic.

The theory of change that emerged for me from this research is that if Claude proactively built desirable difficulties into a dedicated learning mode, the people learning with Claude would be better positioned to achieve real, rather than performative, understanding.

What could that look like in practice? Explain-back prompts, concept retrieval checkpoints that force you to apply what you’ve learned in contexts you haven’t seen, a persistent learning context that remembers what you’ve studied across sessions and nudges you to review. Scaffolding for novices that nudges them towards analogies so that the tool doesn’t just reward expertise.

Right now, Claude is an excellent thinking engine. To become a true learning partner, it has to start forcing us to think harder while we learn, and use what it knows about us to proactively connect unfamiliar concepts to familiar ground.

The Cambridge student was pissed because the test showed that learning with AI hadn’t helped him as much as he thought it had.

That gap between the feeling and reality of competence is one of the most important UX problems in AI right now. But it’s also one of the most solvable.

What practices have you adopted that help you learn effectively with AI?

This week’s content catalog

• 📚: Siddhartha by Hermann Hesse (this is part of taking a class this quarter where we get to read a book a week and discuss how we want to show up spiritually/morally in business and in life.)

• 🎶: ‘Kehlani’ - Kehlani’s latest album (it’s crazy how half the reason I love Kehlani as much as I do is because a friend I love loves her so much.)

• 🎬: BEEF (Season 2) on Netflix (I just watched BEEF for the first time last week. Don’t crucify me. But, yes, I completely get the hype!)

I’ve often wondered what it would feel like to speak directly with my past ‘selves’ in a way that allows a real exchange of insight across time.

What would the 19-year-old me say about some of the questions that are holding 27-year-old me in a chokehold? What patterns would he recognize that I might be too overwhelmed to see right now?

I was able to materialize that by building a digital portal from a corpus of more than eight years of entries from a personal journal of quotes recorded from thoughts, conversations, overhead remarks, films, books, and songs that helped me interpret the situations I was navigating at the time.

Over 500 of these entries were originally written in a small notebook I’ve carried with me since 2018, recording things that were relevant to some questions I had, even if I couldn’t yet articulate their weight.

Carrying the notebook around worked fine until parts of it started to get ruined as water spilled on it and almost erased years’ worth of personal archives.

Because I don’t always have the book on me, it’s also impossible to consult it at the exact moment I want to. And even when I do find it, searching for the most relevant quote is its own mini‑odyssey as I have to sift through hundreds of entries before landing on the one that speaks directly to my question.

I’ve come to see these sorts of interactions with personal troves as a form of temporal wisdom: guidance from perspectives recorded across different moments of one’s life.

Rather than relying only on the limited vantage point of who we are today, temporal wisdom emerges when past reflections and present experiences meet, creating continuity between who we once were and who we are becoming.

At first, the physical notebook was a private ritual of building a personal wisdom bank and preserving the nuggets I collected over the course of my life to, at some point, potentially weave into my memoir—if I live long enough to write one. But as the entries piled on, I realized I was unintentionally constructing something larger than a ‘compendium of wisdom,’ as I’d titled the notebook.

What I was building was a cathedral of perspectives archived over the course of many years, with each quote I recorded tied to a specific moment.

Revisiting those quotes years later felt less like reading pretty sentences and more like reading how a younger version of me was trying to, in his own way, build an independent understanding of the world.

Using Claude Cowork, I turned screenshots of pages from that journal into a single spreadsheet of quotes. After 30 minutes with my girls Lovable and Claude Code, I transformed the Excel sheet into an interactive platform hosted on Vercel that I now call my Wisdom Cathedral: a private platform I use to allow my past reflections to actively participate in shaping present thinking.

Let me talk you through how it works.

In this example, I sent the platform the thought:

“I’m torn between leaning into a romantic connection or letting it slowly fade. I’m equally scared of getting hurt as I am of hurting them, as I am of never trying. HELP!”

Typical shower thoughts, you know?

It then searched across my entire journal and generates a response connecting journal entries from different periods of my life into coherent advice for my current self. Rather than offering generic ChatGPT-style advice, it reframes my thought through the lens of my own recorded notes, surfacing connections that I honestly would’ve missed.

It goes a step further by concluding with thoughtful reflective prompt that invites me to interpret the significance of what I am experiencing rather than rushing to move past it.

In this format, the cathedral also functions as a living journaling companion. Throughout the day, as I interact with the platform, it synthesizes my questions and reflections into the “Journal” tab, creating a record of what I was thinking about that day and a summary of the targeted advice I got.

As someone who cares about recording my day but struggles with finding time to sit down and journal, having an auto-generated daily journal has been such a game changer!

My Cathedral’s daily “Knock” feature surfaces an unexpected fragment from the archive as a small ritual of inspiration that often feels uncannily relevant to whatever I happen to be navigating at the time.

The “Insights” feature shows me the broad themes of the wisdom nuggets in my journal as well as a top-line view of some of my most frequently asked questions.

My favorite feature though is “Wrapped”, which aggregates all the questions I have asked over the course of the year and identifies the dominant themes shaping my intellectual and emotional life, allowing me to see not only isolated moments of reflection but the broader arcs of what I have been learning and thinking about.

My experience with my wisdom cathedral showed me that wisdom (the thing I’m most in search of) is earned through deliberate cultivation.

For me, that’s been my 8-year accumulation of thousands of small observations layered over time that allows for patterns to emerge. But without that initial recording, those wisdom nuggets would’ve vanished, and I’d have had to learn those same lessons over again, more times than necessary.

Recording and revisiting those reflections creates connective tissue between my past and future self, consolidating insights from the past so I can make decisions in the present with extra context from my wishes for the future.

Sometimes it surfaces wisdom that’s screaming too loud ngl. But I still take it. I like that it helps me stay consistent with who I said I was going to be.

For the site, I chose the title of “cathedral” to capture the idea that, just like medieval cathedrals, cultivating wisdom takes years of thankless effort to realize. Sometimes even one lifetime isn’t enough.

Cathedrals like the Sagrada Familia were begun by one generation, expanded by another, and sometimes finished centuries later by people who would never meet the original architects. They did it just for the sake of continuing what was started.

Each group of builders contributed stones with the understanding that they were participating in something that extended beyond themselves.

I think personal knowledge systems work in a similar way.

Every recorded note or reflection is a stone placed deliberately into a structure that will only fully reveal its meaning or value when it attaches itself to a future-state context beyond the grasp of our present selves that’ll breathe new meaning into it.

I now treat documenting as a long-term architectural design practice that helps me build the intellectual scaffolding to support the thinking of my future self.

AI is truly the accelerant for me here.

For the first time, I was able to go beyond cataloging years of personal reflections to being able to also interact with them, query past thinking, surface forgotten insights, and compress years of trial and error into a reflective conversation.

Building my wisdom cathedral got me excited about the potential of AI to transform personal archives from static repositories into active collaborators in the process of self-understanding and meaning creation—two critical levers for combatting the ‘four horsemen of despair’ (which I wrote about here).

To savor life, then, is not only to live it fully but to also preserve it thoughtfully.

What cathedral are you going to build and what stones do you need?

I wrote the first draft of this essay for ‘AI for Human Flourishing’, a class I took in my 2026 winter quarter at Stanford. It’s been edited to sound less academic (used here as euphemism for “stuffy”).

This week’s content catalog

• 📚: ‘Thinking in Systems’ by Donella Meadows (cos I think systems thinking is going to be one of the most cardinal skills of our generation. More on this in another note.)

• 🎶: ‘BULLY’ - Kanye’s new album (is ‘separate the art from the artist’ still a thing in 2026?)

• 🎬: Fate/strange Fake on Crunchyroll - it’s a new anime I didn’t think I’d enjoy, but it’s been giving me life lately.

• 💡: I’ve been following OpenAI’s nine-figure acquisition of the TBPN podcast and what it could mean as a new template for tech-media partnerships as well as the risk of a new format of disseminating AI propaganda (at a time when we really could use less of that). I have my thoughts, but time will tell which it’s gonna be.

This shouldn’t be new information, but I’m Nigerian.

What may be new information to you is my tribe; I’m Igbo.

Anyone who’s read Nigeria’s most famous writer, Chinua Achebe, shouldn’t be a stranger to the Igbo tribe or the word chi. For us Igbos, chi means ‘spirit’.

The most powerful spirit, we call Chineke, meaning the ‘spirit of creation’. It represents the supreme deity, often interpreted as a dual force of thought and action, as creation is what happens when both forces intersect.

Defining ‘God’

The Igbo understanding of the supreme being is not as a ruler sitting on a throne in perpetuity, but as the all-knowing force that fuses thought and action. For us Igbos, God is just as much a noun as He’s as a verb.

Creation, happening.

The book of Genesis shows Christians something similar as God’s first act is manifesting the thought: ‘Let there be light.’ The resulting action, an explosion of light which scientists explain to us as The Big Bang, is believed to have happened at the same instant as His thought. He then spends the next few days creating the universe with nothing but His thoughts, like a vibe-coder today barking requests into Claude Code.

Except that, for God, thought and action are simultaneous.

Now, I know the parallels aren’t airtight. A physicist will likely say that the Big Bang wasn’t instantaneous but unfolded over billions of years, and a theologian might insist that God’s ‘thinking’ isn’t quite the same as cognition in the human sense. Fair.

But the convergence is still striking as, across traditions, the highest conception of a creator is one for whom the distance between thinking and doing is zero, representing the pinnacle of omnipotence.

That’s the definition I want to work with for now.

Our History of Creating God

If we are to ‘create God’ (as blasphemous as that may sound), then we must build tools that bring the distance between our thoughts and our actions closer to zero.

We’ve kinda been doing exactly that, for centuries with technology.

Every technology humans have ever invented can be understood as a device for compressing the thought-action distance. The printing press compressed the thought of ‘I want to tell someone far away what I’m thinking’ from weeks to days. The telephone compressed it further to minutes. Then the mobile phone and the internet shrank it to seconds. Each invention found a gap between wanting and having, between thinking and manifesting, and closed it.

But those were all simple thoughts paired with simple actions. Wanting to hear grandma’s voice. Wanting to see a friend’s face. Wanting to know what the weather will be tomorrow. One thought, one action, little complexity. We’ve already solved those.

Today’s technology lets us turn messy, complex thoughts into a few simple actions that, taken together, create outsized outcomes. But the next frontier is technology that allows us to perform complex actions directly from our complex thoughts without any translation or simplification.

For example, wanting to get the unpublished Harry Potter manuscript found, printed, bound, and in your kids’ hands before their train pulls out of the station to grandma’s. (Ten points if you caught the ‘The Devil Wears Prada’ reference). That thought contains a dozen sub-actions, each requiring coordination, context, judgment, and the ability to manipulate the physical world.

While we’re not quite there yet with AI, what makes the current moment feel different from every previous technological leap is that we’re not just building another tool. We’re building a tool that makes other tools.

Psalms 82:6.

Ye are gods, children of the most high.

If that Bible verse (one of my faves) is right and we are indeed godlike, then the compounding of human progress with each generation inevitably moves us closer and closer to the god-state of having null thought-action distance.

Maybe that’s what we talk about when we talk about AGI: the moment when we create an entity capable of performing complex actions at the speed of its complex thoughts. Although the digital thought-action gap is fast-compressing thanks to the dance between our gadgets, the internet, and AI, the gap in the physical world hasn’t quite budged.

While AI can prepare your fancy product images in a minute, it can’t yet manufacture, test, rework, and ship the product on its own. Second, so long as we interact with AI through an interface, whether a keyboard, a screen, or even our voice, there’ll always be a thought-action lag as we’ll have to translate our thoughts into words, then into keystrokes, then wait for a response.

This leads me to an uncomfortable conclusion.

The Cost of Creating God

To truly create God, to collapse the thought-action distance to zero, we’d have to put the tools for complex action directly in the seat of complex thought: the brain. No interface. No translation. Thought meeting action at the speed and site of its formation.

To create God, we’ll have to create the singularity—that dreaded fusion of mind and AI, of man and machine. The apex earthly force of thought married to the apex earthly force of action.

What keeps me up at night is whether the distance between thought and action is itself what makes us human. And if, by closing it, we erode something at the core of our humanity.

What if that thought-action gap isn’t a problem to be solved, but a needed space for us imperfect thinkers to weigh the consequences of our actions, and choose differently?

Creating God requires removing humanity entirely from the decision-action loop that follows our thoughts because our deliberation slows down speed, efficiency, and agency—the three bastard demigods of the capitalist machinery upholding our societies.

Every spiritual tradition, including the Igbo one, teaches that the distance between thought and action is where morality lives. It’s where we deliberate; where we exercise judgment; where we get to choose to become more than just rule-following bots.

Chineke can afford zero thought-action distance because His thoughts are supposed to be perfect. Ours are not. “Ye are gods,” yes. But young, reckless ones.

With every new invention marching us closer to that single fated singularity moment, sooner than later, we’ll need to reckon with what it means to wield godlike powers of zero thought-action distance before acquiring the godlike wisdom needed to not self-destruct.

So the real question isn’t whether we can create God (we’d likely get close), but should we?

Thanks to Jeremiah Nnadi and Temitope Stephens for reading earlier drafts of this.

A few weeks ago, I was at Anthropic’s 1st year birthday party for Claude Code.

Naturally, the room was packed wall-to-wall with some of the most hardcore fans of Claude and a swathe of the Anthropic employees (including Boris Cherny, the creator of Claude Code).

It shocked me to find out that that was the first public event Anthropic had ever organized to directly celebrate a Claude product. But even that event didn’t feel as much of a fan club for Claude and Anthropic as the last few days on X have felt.

Anthropic (the AI lab behind Claude) has been getting flowers with everything going on right now in Washington and for the near-flawless performance of Claude (we’ll kinda forgive yesterday’s blackout 🫠), while also getting dragged for its high prices.

The aura-maxxing Anthropic has been getting over the past few weeks, including from its hard line on ethical AI use, has had me thinking a lot about how Anthropic’s brand strategy feels straight out of the playbook of Beyoncé Giselle Knowles-Carter.

The near‑mythic stories about Claude’s elite performance, Anthropic’s willingness to take public political stances, Claude’s exorbitant pricing, and the CEO’s careful rationing of visibility all line up with Beyoncé’s brand DNA of technical excellence, premium positioning, clear politics and controlled scarcity.

But why does brand even matter in AI? And why am I comparing a foundational model to the most decorated artist in GRAMMY history?

On my Caltrain ride from campus to Claude Code’s birthday party, I was reading this essay, Tokens Aren’t Fungible, written by Nnamdi Iregbulem, a partner at Lightspeed Ventures. I’d been trying to connect with Nnamdi for months and wanted to catch up on his writing before meeting him at the party.

I found the essay’s main argument—that past a certain performance threshold, the brand of an AI lab is what drives the lion’s share of how much consumers are willing to pay for its models—very interesting.

Nnamdi argued that models are not commodities priced purely by intelligence; the identity and perceived qualities of the lab doing the shipping explain more of market reality than the model performance benchmark charts do.

For AI labs, brand isn’t just a decorative artifact. It’s the critical moat.

Claude has gotten it mostly right so far, but it risks drifting into a kind of self‑inflicted inaccessibility if some of the very features that make it feel premium end up also keeping people out.

As one of Anthropic’s ambassadors at Stanford, I’m naturally one of the biggest Claude fans. But as a power user, I’m one of its most unrelenting, constructive critics. My goal in this essay is to use the current hype cycle only as a backdrop for dissecting Claude’s unique approach to winning the hearts and wallets of AI consumers.

Let’s get into it together.

Premium Brand • ‘I’m That Girl’

Claude’s Super Bowl ad was the clearest signal to me that Anthropic understood something most tech companies don’t: luxury brands compete through distancing. You don’t sell the product. You sell the distance between those who own your product and those who don’t.

Earlier this month, Anthropic spent about $30 million on Super Bowl ad slots for its tongue-in-cheek ad that mocked the idea of ads inside AI chatbots. The message was blunt: “Ads are coming to AI. But not to Claude.”

It’s the same move Beyoncé made with “Formation” and the 2016 Super Bowl halftime performance using the biggest stage platform to declare publicly what she stands for: Blackness, feminism, Southern culture, racial justice, and social commentary. The exception is that she got paid while Anthropic paid, but let’s not squint too closely at that right now.

The point is you didn’t have to agree with her but, on that stage, you couldn’t ignore her. Instead of selling features, Anthropic sold differentiation through moral distancing: our competitors are willing to ruin your AI chatbot experience with ads, but not us.

According to CNBC, that ad led to an 11% spike in Claude’s daily active users. TechCrunch noted the Claude app climbed to No. 7 in the App Store shortly after the Super Bowl broadcast (it’s now No. 1 on the App Store).

Even Sam Altman’s reaction to the campaign, whether intentional or not, amplified Anthropic’s message, further cementing their differentiation from OpenAI’s announcement on bringing ads to ChatGPT. I personally think that when your competitor’s CEO is publicly annoyed about your marketing, your positioning has done its job.

Unlike ChatGPT, Claude isn’t trying to be the AI for everyone. It’s instead positioning itself as the AI for people who care a lot about performance, safety, and, dare I say, human flourishing.

It’s the same way Beyoncé doesn’t look to be a mass-market pop artist that dominates all the charts (no, this isn’t an excuse for her weaker charting compared to some of her peers). She focuses squarely on audiences that care about artistry, technical excellence, cultural legacy, and political commentary. She sets the price that matches the quality of the performance she consistently delivers and she waits for the rest of the world to meet her there.

In economic terms, both Claude and Beyoncé function somewhat like what economists call Veblen goods where their high prices themselves feel like part of the value proposition. You’re not paying despite the cost; you’re paying because the cost signals status and promises a premium experience.

With both charging $20/month for their pro subscription, Claude’s user base of roughly 19 million might look like a weakness compared to ChatGPT’s 800 million. But Anthropic is already generating $5B in annualized revenue on that much smaller base, with a run‑rate that’s rapidly closing the gap with OpenAI’s $12B. On a per user basis, that suggests Claude’s customers are both using it more heavily and are willing to pay more for what they’re getting compared to ChatGPT's.

Claude’s premium positioning comes from focusing on a smaller user base that places a premium on model quality with a willingness to pay a price closer to true compute cost, while ChatGPT targets an enormous user base through aggressive subsidies to cater to their lower willingness to pay.

Political Positioning • ‘Okay, ladies, now let’s get in formation’

Every AI company claims to care about safety and ethics, but Anthropic has made it its entire personality. Claude isn’t just an AI assistant; it’s an AI that wants you to know it has a conscience, or at least a constitution.

Now, I can get into all the reasons why I think centering a corporate brand on ethics is a brittle brand position in corporate America (I’ll save that essay for another day), but it seems to be working for Anthropic.

Their technical framework, Constitutional AI, is a public values statement that makes it clear beyond their engineering approach, the philosophical foundations behind how they think about AI. Claude even has a publicly available constitution that’s essentially a set of red-line principles that it’s trained to follow no matter what.

Same playbook as Beyoncé. She doesn’t just talk about what she cares about, she dedicates entire albums, visuals, and tours as temples to them.

‘Lemonade’ was a statement of relationship politics braided neatly into a pop album. ‘Formation’ was one of Beyoncé’s most rule-breaking performances that she did in service of her politics. The Super Bowl halftime show has a long history of discouraging political performances as the NFL believes that sports should remain an apolitical arena to be able to connect fans from all walks of life.

With her ‘Formation’ performance that roared into the stadium with hundreds of Black dancers dressed in Black Panther outfits, Beyoncé looked at the NFL’s rules on politics and said ‘cool story bro’.

Not even a stage as grand as the Super Bowl, arguably the biggest stage in music, could stop Beyoncé from wearing her full armor and being battle-ready for the sake of her politics and using her privilege and platform to stand up for the identities she represents.

Anthropic’s position on AI ads and AI safety is its Formation moment.

With one ad, Claude established its premium distancing and also reinforced its principled stance on AI.

While every other tech company used their 30-second spots to sell features, Anthropic had no product demo at all. Instead, its ad was a declaration of values disguised as a commercial that was essentially a short manifesto on how Anthropic thinks of responsible AI.

Last week though, TIME reported that Anthropic has backtracked from their flagship pre-deployment safety pledge where they promised to stop developing new models if they can’t pre-guarantee the adequacy of the company’s safety measures. On one side, you have a lab positioning strongly as the antithesis to irresponsible AI design and, on the other, a company softening a public promise about how carefully it will ship its powerful models because its competitors aren’t making the same commitment.

That tension is the crux of the story of building a premium AI brand in 2026 America.

Anthropic is trying to be both the principled lab that says no to certain forms of capitalism and the ambitious company racing to ship state-of-the-art models in a brutally competitive AI market. Just like Beyoncé is billionaire business mogul and voice of the resistance in the same breath, Anthropic is selling us safety and scale simultaneously. Anthropic’s brand strength comes from embracing this tension rather than pretending it doesn’t exist.

They do this because they can’t be one without the other.

Beyoncé needs to be powerful within the establishment to break the rules without much consequence and speak truths that would end the careers of less influential artists. It’s the same for Anthropic. The company needs to be at a massive scale to build competitive products that provide a strong enough counterweight to OpenAI.

Despite its caution around the militarization of AI, Anthropic still needs to collaborate with the government to earn the revenues required to grow. Politics are part conviction, part choreography, and that’s exactly what makes Anthropic’s “luxury ethical AI” positioning so compelling, and also so fragile.

Performance Excellence • “I grind ‘til I own it”

All the premium branding and political positioning in the world doesn’t matter if the product doesn’t perform well. And “perform well” sits at the center of what both Beyoncé and Claude are known for. In fact, this is where their arc reaches its crescendo.

Anthropic’s Claude Code went viral in the early days of January when a principal engineer at Google publicly stated that she had “recreated a year of work in an hour” with Claude Code (not Gemini btw). Weeks later, we saw Jensen Huang, CEO of NVIDIA, casually mentioning Claude in earnings calls and advocating that every software company needs to use it.

According to a Blind survey, 31% of US tech workers now use Claude as their primary AI tool. 50% of Meta employees and 34% of Microsoft employees are choosing Claude over their own company’s Copilot, which only 32% of employees preferred.

The jokes really do write themselves, don’t they?

Anthropic’s most powerful model yet, Opus 4.6, dropped last month with a mix of fanfare and dread as people suddenly saw what was possible with that much better reasoning. A few weeks later came Claude Cowork, an on-device AI partner that serves as connective tissue between your local machine and the rest of the digital world, letting you build and interact with simple prompts. It kickstarted conversations in elite tech circles about what the future of work looks like when we all have a capable AI collaborator sitting on our desktops.

And here’s the other thing: Anthropic almost never does teasers. Certainly no teaser tweets from the CEO and no countdown timers. They just drop it like it’s hot (as it almost always is).

This is the same approach Beyoncé perfected by letting her work speak, minimizing pre-launch noise, doing almost no interview, and trusting that the quality of her artistry will generate its own inevitable press cycle. This has been the case for every new drop from Anthropic over the past few weeks.

Anthropic’s CEO, Dario Amodei, also doesn’t do tweet-storms like some others. Instead, he communicates with the world mostly through long essays like his iconic Machines of Loving Grace and more recent The Adolescence of Technology— the kind that take 45 minutes to read and get cited in industry forums and reports.

It’s the opposite of Sam Altman’s strategy, which seems like being everywhere, all at once, on every podcast, at every hackathon, and in every X thread. Neither approach is wrong. But one of them builds a mystique that crystallizes into cultish intrigue. And I don’t know about you, but I’ve never seen anyone build a cult following around excessive accessibility.

Where Claude’s Beyoncé Strategy Falls Short

So, inasmuch as I’m an ambassador for Claude, I’m not blind to its shortcomings. From performance to positioning, I have a laundry list of notes on how Claude can be and do better. But let’s start with some of my most pressing:

1. ‘Premium’ can quickly become ‘inaccessible’: The greatest risk of premium positioning is that you could end up so out of touch that the average customer has never heard of you. You can be Beyoncé in the AI labs, but if the average person on the street doesn’t know your name, the brand ceiling is real. Beyoncé is, after all, still Beyoncé in part because billions of people know who she is. “Premiumness” without the consumer awareness to match is obscurity with lipstick on (I’d know, cos I previously ran a magazine that kinda fit the description).

2. Being hyper-competitive can give way for hypocritical behavior: Dropping the pre-deployment safety pledge while simultaneously running ads about how much you care about responsible AI is tense, if we’re being generous. Anthropic’s safety-first narrative only sticks if the company’s corporate behavior stays at least directionally consistent with its wider marketing. With all of its industry-shaking product releases and the goodwill from consumers, Anthropic has more credit to spend here than most companies. But goodwill is a depreciating asset, and every future gap between messaging and action will only fracture the trust the company has built with consumers.

3. Premium brand positioning can sometimes oversell your product experience: New customers switching from ChatGPT to Claude are in for a few surprises. While Claude is a superior product by almost all accounts, ChatGPT still has advantages in chat context window length. The frequent rate limits on Claude are one of the most common complaints I’ve gotten from members of Claude@Stanford, one of the premier clubs of student AI builders at Stanford. Yesterday’s Claude blackout almost caused a protest in our WhatsApp group lol.

   

   
   While some of this might be indeed unavoidable technical failure, it’s tricky when it contradicts the brand positioning around a more premium experience.

What Claude’s Playbook Taught Me

Whether or not Anthropic consciously modeled its strategy on luxury branding (and I suspect they didn’t, which made it more interesting to me) I found five principles worth stealing:

• Pick a villain. Unlike ChatGPT, Claude has clear villains: ads, human surveillance, fully autonomous weapons. You don’t have to agree with them, but you know for sure what they are.

• Make your values loud and expensive. Anyone can write a mission statement, but Anthropic is willing to walk away from $200M defense contracts, which puts a price to their values. Not many companies can pay that price.

• Let the product be the press tour. Claude Code didn’t need a launch event. They only had one to celebrate its one-year birthday, which I was lucky to attend in SF a few days ago. The best marketing is a product so good that users become evangelists. Claude has that.

• Court the tastemakers before the masses. Anthropic focused on winning over developers and power users first. They did the harder thing first and the wider consumer market followed. This is Beyoncé playing at Coachella before announcing a stadium tour, earning the approval of her most engaged audience, before letting the mainstream in on the fun.

• Embrace tension. Claude’s brand, like most luxury brands, contains contradictions. But these tensions make the brand feel alive, because they’re the same tensions any organization navigating real stakes will face.

I honestly can’t say whether Claude will still retain this moat in five years as models keep leapfrogging each other. A few months from today, some other lab might release something that makes Opus 4.6 look quaint.

Thinking back to that evening at Claude Code’s birthday party in SF where a room full of builders celebrating a command-line tool like it was an album drop, I realized no one was there because of a benchmark chart.

They were there because Claude Code had changed how they worked (the winner of the birthday hackathon was a personal injury lawyer who’d never written a line of code), and just wanted to be in the same room with the people who built it. “Claude Code changed my life” as another hackathon winner put it. And that’s what differentiates a product from a brand.

If Claude and Anthropic can retain this brand moat for as long as Beyoncé has held her reign as the most decorated woman in Grammy history, they could just be crafting the beginning of a century-defining company that becomes culture.

All of this matters because the AI model wars won’t be won by the smartest model. It’ll be won by the lab people believe in.

It’s now up to the different labs, from Anthropic to OpenAI and everyone in between, to give people something to believe in beyond bringing AGI to life.

My next essay will summarize research I conducted last quarter in an education technology class at Stanford to test how effective Claude is at helping people deepen their understanding of complex concepts. The findings were surprising to say the least.

If you enjoyed this, you’ll want to be subscribed for that one.

Thanks to Jeremiah Nnadi, Adebayo Adegbenro, and Hope Moussi for proofreading earlier drafts.

I spent the past week building my personal website.

It’s taken years of procrastination and mulling over whether I should even do it.

It always felt vain and unnecessarily expensive to have a personal site until a few recent cultural shifts made me completely rethink their place in the future of work.

First, more employers are demanding personal websites and portfolios.

If you’ve applied to jobs in the last few years, you’ve definitely noticed this creeping trend. Lodged somewhere between the field for resume, cover letter, and LinkedIn URL is the “personal website/portfolio” form field. Sometimes optional, but increasingly not.

It’s kind of like what happened with the LinkedIn URL field in the 2010s. It snuck its way into application forms at a time when social media was taking off and having a LinkedIn profile was a way to signal both a professional brand and social media fluency.

Although not every application form today has the LinkedIn URL field, leaving it empty still leaves the uncomfortable feeling of leaving a stone unturned in a fiercely competitive job market where standing out of the crowd has never been more important.

But LinkedIn started to become insufficient when hiring managers started looking for more individuality in a sea of sameness from candidates applying with identical backgrounds. There’s only so much you can optimize your LinkedIn profile to stand out from other LinkedIn profiles. And this is where personal sites come in.

Personal Sites as Fluency Portfolios

With the tools for creating them now readily available thanks to AI, personal sites in 2026 are just as much about communicating technical fluency as they’re about communicating professional identity. Today, when AI fluency is at a premium, having a personal site built by AI tools (which takes an afternoon) allows you to construct your own ‘digital living room’ to present your profession and your craft cohesively.

LinkedIn’s 2025 AI Labor Market Update reported that job postings requiring AI literacy skills are growing at more than 70% year over year as of early 2026. When a recruiter visits your site and sees something personal beyond what a paid developer could produce, they see both your personality and proof that you can collaborate with AI to produce something real.

In a job market where everyone claims AI skills but few can demonstrate them, a personal site becomes the portfolio for AI fluency, the most important professional skill of the decade.

Personal Sites as Identity Anchors

I learned while journaling recently that your profession and your craft are two very different things. Your profession is what you’re paid to do while your craft is what you can’t help but do.

Communicating your profession is easy; there’s CVs and LinkedIn for that. But communicating your craft? Now, that’s much harder, especially when it doesn’t fit neatly with your profession.

Take an example from Nnamdi Iregbulem, one of Silicon Valley’s most prominent Black investors and a personal idol.

His site opens with a clear declaration of his identity at the top: “venture investor and writer”, embracing both his profession and his craft while his LinkedIn header which reads “Partner at Lightspeed Venture Partners” leans exclusively on his profession.

You could likely tell that he writes from the fact that he shares some of his writing on LinkedIn, but his site lets him claim that identity cohesively in a way that also communicates his personal taste.

Personal Sites as Taste Signatures

When I sat down to build my personal site last week, I didn’t write any code.

I simply described what I wanted in natural language. I wanted something that felt like walking into someone’s home, but online, where one leaves feeling like they’ve just spent time getting to know a real person. So I whipped out Claude Code to start building.

While Claude handled all of the execution, I owned all of the taste: choosing the graph paper background; selecting the symbols (◆, ▲, ✦) as a visual motif; picking a color palette; defining the tone of voice; and deciding on the information hierarchy.

I kept a ‘devjournal’ to log my prompts (almost 3K words of them) and micro-decisions I made across the 21 different versions Claude coded up for me. I wanted to be able to reflect on the efficiency of my prompting and the overall build process when I asked Claude for the build report, which I always use to learn how to build more efficiently next time.

This is the underappreciated dimension of AI-built personal websites: they’re taste tests. Two people can use the exact same tools, spend the exact same amount, and produce wildly different sites because the quality of the output is a function of the quality of the creator’s instructions.

Your website reveals what you notice, what you value, what you find beautiful. It reveals whether you have an eye for detail or whether you’ll accept the first thing the machine spits out. In a world where AI handles execution, the ability to discern good from great becomes the key differentiator. Your personal website becomes a portfolio of your taste.

Take an example from a personal website like this one by an Antoine Wodniack.

Every element of his site communicates that this was built by a tasteful engineer, so even without the ‘Creative Developer’ tagline on the homepage, we already get a sense that his site is the digital living room of someone who’s just as creative as they’re technical.

But there’s another reason to build one I hadn’t considered.

Personal Sites as Owned Archives

I got to meet Nnamdi in person for the first time yesterday and I asked him why he chose to build his site versus keeping his profile and writing mostly on other sites.

He said that he decided he wanted to pool together all his writing work to a domain he owned. He wanted to maximize his digital autonomy while creating a more permanent record for the evolution of his thinking, as a lot of these publishing platforms come and go.

He was right. So much of my writing is littered all over the internet (some on platforms that are now defunct). My early WordPress days in high school gave way to LinkedIn and Medium around 2017, then Beehiiv and Substack around 2020. I’m only now starting to centrally archive all my published work to make them accessible in one place.

That’s the next build I’m focused on for my personal site, especially as my thinking on them evolves from ‘landing pages for successful people’ to ‘personal digital living rooms’ that lets people set their own rules for communicating their backgrounds, identities, tastes, and skills with the world.

In a world where the personal site field is increasingly standard in applications, you’ll eventually have to decide—personal site or blank box?

But the real question isn’t whether you can afford to build one, it’s whether you can afford what leaving that ‘personal website’ field blank does to your prospects as the job market becomes more competitive.

If you’d want to build one in an afternoon, even if just for fun, I’d recommend getting started with Claude Code, as it can help you whip one up with just $20 and a few focused hours. This tutorial by Nate Herk is a great place to get started.

If you do build (or have) a personal site, please share with me in the comments! I’d love more inspo on how to improve the furnishing of my digital living room.

Two days ago, I stayed up until 4am building apps on my phone...as one does.

I had a 9:45am meeting and was exhausted out of my mind, but I was also drunk on a very specific kind of joy I thought I’d lost as a kid.

You see, I grew up a LEGO kid.

Wasn’t great, but I loved my LEGO blocks. Like most kids, I built everything with them: cars, tables, houses, more houses, and ofc some unidentifiable structures.

I loved building things simply because I could. Also, watching ideas in my head take form by my hands, LEGO block by LEGO block, gave me a needed sense of agency as a kid in an adult world.

The sense of play that comes from building on my own terms without the weight of perfection is something I’ve been trying to reconnect with, and I’ve made good progress the last few weeks thanks to some AI tools I’ve been playing with.

With AI, we’re living through a once-in-a-lifetime shift in how we experience technology as we go from personal computers to personal software—tools built not for millions of users, but for one.

AI is democratizing software development so fast that Stanford created a new course called ‘The Modern Software Developer‘ to teach no-code students how to build with AI and CS students how to 10x their capacity.

What’s fascinating about this shift is that for decades, we optimized for scale. One app for everyone, for everything. But AI lets us flip that model. Instead of us building for billions, we can build for one.

Our software can now become as personalized as our playlists, our feeds, our entire digital lives.

We’re not just consuming personalized experiences anymore, we’re creating them.

How a Stanford Class Tickled My Inner Tinkerer

One of the classes I’m taking in business school this quarter is called ‘AI for Human Flourishing‘, taught by the legendary Professor , and it’s helped me a lot with reconnecting with my core craft of creation.

The class is as much a building class as it is a tech-philosophy class, but it isn’t about building the perfect product or the next unicorn. Instead, it’s about using AI however you see fit to reconnect with whatever makes you flourish (however you choose to define that for yourself).

For our ‘buildathon’ in class, imperfect prototypes were considered a success because improving was the point, not proving. That was refreshing.

Through that buildathon, we got early access to Wabi, a platform currently in invite-only private beta that’s already being hailed as “the YouTube of apps.”

Founded by Eugenia Kuyda (a founder I admire who previously built Replika, the pioneering AI companionship app), Wabi is a personal software platform where you can create mini apps using nothing but prompts. You don’t need to know how to code. You don’t need to understand databases or APIs or deployment pipelines.

On Wabi, you just describe what you want in a chat: “build me a mood tracker with energy logs morning, afternoon, evening, and night” or “build me plan my weekly meals based on what’s in my fridge”—and Wabi generates a fully functional app, complete with interface, logic, and even an icon.

Well, Wabi calls them “mini apps” as they live inside the Wabi app to solve specific problems rather than be everything to everyone.

To edit the app, you just need to chat with it in natural language about the changes you want made, it’ll give you a time estimate of the edit, and will ping you once the app’s been updated to your specifications.

But what I find special about Wabi is that it’s social. I can browse apps other people have built, ‘remix’ them for my own needs, comment on them, and even share them with friends. It’s kind of like if YouTube let you not just watch videos but also remix them, change the ending, add your own scenes.

Some of My Personal Apps

Before Wabi, my portfolio of personal apps was mostly limited to Claude artifacts like the Pomodoro timer I built to play Hedwig’s Theme when time runs out.

While it’s great, I was limited by it being web-based.

Since getting access to Wabi, I’ve become ‘Bob the Builder’ of Wabi mini-apps that live in one app on my phone saving me storage and giving me the agency to create my own versions of my app, as many as I want, whenever I want.

Some of my favorites personal apps are:

Guess the Word game: With this, I just wanted to scratch an itch for a game that let me practice both my vocabulary and recall, while reminding me to stay stocked on avocados. I had one of my friends play it with me and he loved it too, before sending an exhaustive list of things to improve with the gaming experience (such as more fruits than just avocados). To work, I went.

Meme Generator: My ‘Toasty Memes’ mini-app was to solve the paralysis that descends on me whenever I have to search for a meme to include in my broadcast to my class mailing list as ‘blast tax’ for spamming their inbox. Yup, that’s a real thing here. Now, I mostly use Toasty Memes to crack myself up at both really good and really bad memes.

Gamified Calendar: I made the ‘Calendar Go’ mini-app, where I win XP for completing tasks and events on my Google Calendar, as a more fun way of going through my day. I also wanted to use earning XP for tasks as a way to incentivize myself to time-block my daily to-do list. It worked...for a while. I lasted four days before I realized I’d gamified my life so hard I started over-scheduling just to farm XP. So I stopped.

Check my page on Wabi for my full list of apps.

I’m still not that good but Wabi lets me build and edit apps tailored to my exact needs with my furiously-typed prompts.

I’m excited about personal apps and how platforms like Wabi are helping make it easy for anyone with a smartphone to experience the joy of building digital tools for themselves and their loved ones.

No extra in-app purchases and ads, just a world of your own creation waiting for you to step in.

How many times have you thought “I wish there was an app for that“ but it was too niche for any app developer to care about? How many small annoyances in your daily routine could be solved with a tool built exactly for you?

That gap between what exists and what you ACTUALLY need is where personal apps live.

Filling that gap for myself has helped me create moments of ‘flourishing’ that has translated physiologically into extended periods of flow that my Ōura ring picked up as long periods of restoration.

Beyond the utility of personal apps, building them alone restores me.

Maybe it’s because they tickle the LEGO-building kid in me—the one who built things not for an audience, but simply because he could.

I’ve found that when I build out of curiosity, failure becomes interesting not devastating. When something doesn’t work, I tinker and tinker until I fix it. Not because I have to, but because I want to for no other reason than “just seeing where this goes”

I’m playing, in the truest sense of the word.

What could be more fun?

What would you build if you could make an app in five minutes? Tell me in the comments; I’m genuinely curious. Also, I have one Wabi access link for the first person who comments with an app idea.

P.S. If you’re on Wabi, add me @heyarinze or via this link.

Current Content Catalog

• 📖: ‘The Anxious Generation’ by Jonathan Haidt - I need to finish this omg

• 🎧: ‘BRITPOP (Deluxe Edition)’ - the new album by Robbie Williams

• 🎬: Bridgerton on Netflix - after leaving S3 in the dust for almost a full year

• 🔎: The Crustafarianism religion created by AI agents on Moltbook – wtf!?

Harry Potter fans need no introduction to the Patronus Charm.

It’s the hardest spell to cast in the Harry Potter universe, not because it is deadly or technically complex, but because it requires summoning a deeply happy memory in the face of abject terror.

It demands blinding light in the exact moment when dark despair envelopes you.

Most wizards can’t produce even a shield form of the charm, let alone conjure a corporeal Patronus (like Harry Potter’s stag, which inspired this story’s cover image).

To conjure a Patronus, a witch or wizard must summon a memory so deeply joyful, so filled with hope, that it pierces through the cold, despair-inducing fog of the Dementors (shadowy creatures that literally kill a person by sucking away all their joy).

Today’s dementors aren’t cloaked creatures gliding ominously over land and lakes; they’re subtler.

Despair is alive in the quiet loneliness we nurse but never name. It’s alive in our disappointment, uncertainty, burnout, comparison, and our constant self-flagellation over whether we’re doing, being, or becoming enough.

Despair is also all around us.

The earth is boiling. Wars are raging. Jobs are evaporating.

The rules of the world we grew up in are changing right under our feet.

Casting a Patronus Charm today isn’t about artificial optimism, it’s about conjuring trained joy in the presence of despair.

If that isn’t magic, I don’t know what is.

The ‘muggle’ version of the Patronus Charm is what I call emotional alchemy, the ability to turn any emotion/experience, no matter how dark or negative, into something bright and positive.

It’s not about ignoring negative emotions/experiences, but about converting them into fuel, clarity, or resolve.

I’ve found that, like magic, emotional alchemy can become an instinct through practice. It involves nourishing the things that make us feel good so that we can tap into them on days when the dementors of life haunt us.

Practicing our Patronus determines how resilient we’ll be when tormented by what I call ‘The Four Horsemen of Despair’: loneliness, apathy, meaninglessness, and helplessness.

• LONELINESS is feeling deeply disconnected from your identity, community, or purpose.

• APATHY is the resignation that comes from feeling like all your efforts are futile.

• MEANINGLESSNESS is feeling like your future is inconsequential or, worse, empty.

• HELPLESSNESS is the feeling that your situation is never going to change and you lack the ability to save yourself.

I describe these Horsemen as feelings because, stripped bare, that’s all they are.

To fight these feelings in our daily lives, we’ll need to practice our emotional alchemy.

For me, that looks like:

• Going on biweekly hikes to touch grass.

• Keeping a private journal where I name my feelings without fear or shame.

• Keeping a gratitude journal I can return to on days when I need to be reminded of the sweetness in my life.

• Meditating to create some distance between my feelings and me, so I can look at them objectively.

Each of these helps me pick myself up on days when I’m caught in the psycho-emotional trenches dug for me by one or more of despair’s horsemen.

If you watch anime, you already get what emotional alchemy looks like.

In most shonen anime, characters usually transform after tapping into their Patronus—a childhood memory of warmth, a flashback to a mentor’s belief, a friend’s encouragement, an image of someone they swore to protect, or simply someone waiting for them to return home.

It’s that love or purpose they remember that helps them to push past their limits and unlock their most powerful moves.

Just like a Patronus.

What anime understands (and we adults often forget) is that power is rarely unlocked by effort alone, but by remembering why effort matters in the first place.

The lesson here is that life will always dish out despair; some seasons more than others.

But each of us has a well of positive memories, relationships, and dreams that, if drawn upon often and intentionally, can become a powerful psychological tool.

We must practice accessing joy even when it’s inconvenient.

We must build rituals that remind us of who we are.

We must actively collect/record memories that uplift us.

Whatever you can, strengthen the muscles that help you return to hope, joy, and meaning—especially when it’s hard.

Hope, joy, and meaning are the tools for winning the fight against despair, the deadliest dementor of our time.

Find them.

Nurture them.

Master them.

Current Content Catalog

Here’s what I’m currently:

• Reading: ‘The Anxious Generation’ by Jonathan Haidt

• Listening: ‘Very Stubborn’ (EP) - Victony

• Watching: Record of Ragnarok on Netflix

• Researching: Netflix, Warner Bros, and the future of interactive storytelling

One good thing about my second year of business school is that I have no midterms.

That gives me more time to yap with classmates and pursue side quests like representing Anthropic at Stanford as a Claude Student Ambassador.

Doing that role in parallel to my dual education master’s program has given me the space this quarter to think about and document how I use Claude to learn.

I wanted to distill some of the best practices I’m adopting as I think about how I want to engage with these platforms to learn faster and more deeply.

On campus, we talk a lot about how AI is transforming industries and economies. But what fascinates me more is how it’s changing us; the ways we think, learn, and communicate with the world.

This essay is my attempt to make sense of that shift; and to share what’s been working for me in using AI chatbots like Claude to learn faster and think more deeply, without outsourcing the thinking itself.

Using Curiosity to Fuel Your Understanding

With how powerful AI chatbots like Claude are at conducting extensive research, performing complex calculations, and creating multimodal content, they’re practically begging to be used as homework answer vending machines.

But using Claude as an answer bot is a short-termist approach to learning. You limit the full power of the chatbot and, more importantly, the full potential of your mind.

When you use Claude to co-construct understanding through an exchange of thoughtful questions rather than as an answer machine, you eventually start thinking like an expert, not just imitating the thinking of one.

• Weak Prompt: “What’s the answer to problem 3 on my Microecon assignment?” *attaches file*

• Strong Prompt: “I’m trying to understand what question 3 in this Microecon assignment means by demographic dividend. How’s that similar to or different from the same mental model for financial dividends?” *attaches file*

Using curiosity-driven prompts transforms Claude into a friendly thinking partner that sharpens your thinking rather than replaces it. That back-and-forth friction is what makes what you learn stick.

Using Analogies to Sharpen Your Intuition

If curiosity breeds depth, analogies breed breadth. The school term moves really fast, and we don’t always have the time (or confidence) to dive deep into a topic we’re confused about to find mental models from other disciplines that can help us better understand it.

This is where AI chatbots like Claude can really shine as they can leverage their vast corpus of knowledge to draw parallels between distant domains to help translate the same idea across multiple domains.

• Weak Prompt: “Explain chemical equations and how to balance them.”

• Strong Prompt: “I’m trying to understand chemical equations and how to balance them. Can you explain it using parallels from basketball concepts?”

When you ask Claude to explain complex ideas through analogies, you’re teaching your brain to see connections between seemingly unrelated things, which is the essence of creative intelligence. You also make it easier for yourself to understand unfamiliar concepts by looking through the prism of familiar ones.

Using Targeted Practice to Build Confidence

Learning science shows that one of the most powerful ways to learn is through error analysis (reverse engineering what you got wrong and correcting it from first principles) and the way to keep the wheels of learning turning is progressive overload (slowly increasing the difficulty of your training).

Claude excels at this because, from your responses, it’s able to thoughtfully explain where you went wrong and how you can get back on track. It can also adapt the difficulty of the training content it creates as your performance improves.

• Weak Prompt: “I don’t get factorials. Can you explain it to me?”

• Strong Prompt: “I just started learning about probabilities. I get XYZ so far and how it applies to ABC, but I’m stuck on factorials. Can you help explain it simply and in a way that builds on what I already know so far?”

For me, confidence comes from familiarity, not certainty. With each practice that exposes my weak spots, I see that studying isn’t about being right all the time; it’s about understanding why you’re wrong, and becoming a little less wrong each time.

Using Claude ‘Artifacts’ to Personalize Your Learning

Claude’s Artifacts let you create custom tools that are perfectly tailored to how you learn. And the best part is that they can be interactive!

To create an Artifact, you just describe to Claude what you want it to build, what content should be embedded, and how the two should interact. I’ve used Artifacts to create everything from flashcards for understanding NPC intelligence in gaming to an app to teach me the basics of mixology.

Claude’s Artifacts are game-changing for personalized learning. Instead of having concepts passively explained, you can create a personalized learning environment with Claude, built around your unique cognitive makeup and learning preferences.

Using Claude ‘Projects’ to Create Your Learning Environment

While Artifacts let you build custom tools for individual sessions, Projects take this personalization even further by creating a persistent learning space across multiple conversations

Think of a Claude Project as your dedicated learning space. You can set up instructions that shape how Claude teaches you whether that means being more Socratic, asking you to attempt problems before revealing solutions, or adapting to your specific learning style.

Claude Projects is designed to help you learn smarter through:

• One-shot learning instructions: By setting your learning goals once while setting up the project, there’s no need to remind Claude at the start of every chat whether you want long/short explanations, prefer being prompted with questions/analogies, etc.

• Consistent reference files: When you add knowledge to the project by uploading textbooks, lecture notes, problem sets, or any other stuff, Claude references them across all conversations within that project, creating a cohesive learning experience.

Projects also work really well now with Claude’s memory feature, creating a learning environment that alongside you while maintaining the structure and resources you need.

Subscribe now

Key Insight: Prompting is the New Thinking

I’ve found that when I write a detailed prompt that explains what I know, what I don’t know, what I’ve tried, and what kind of help I still need, I’m not just chatting with Claude, I’m actually clarifying and extending my own thinking.

The process of drafting the prompt itself is a form of metacognition (thinking about my own thinking). The thing with weak prompts is that they’re vague; and they’re vague often because the thinking behind them is vague.

I’ve come to believe that good prompting is just good thinking made legible.

Good prompts are specific because the thinker has done the work of diagnosing where they’re stuck and what they need to get unstuck within the boundaries of what they’re working on. That kind of engaged learning is what transforms rookies to pros in just a few hours of smart AI prompting.

From observing what prompts give me the most insightful responses from Claude, I created a personal prompting framework called TUGS to help me get better answers and deeper understanding from AI tools like Claude.

• TASK: What am I working on?

• UNDERSTANDING: What do I think I already know?

• GAP: Where exactly am I stuck?

• SUPPORT: Do I need questions, examples, explanations, practice, debugging?

The other reason I named the framework TUGS is that, for me, prompting (especially with Claude) feels like a tug-of-war with my mind.

It’s a whole new way of learning for me, where it feels like I’m actively constructing knowledge by wrestling the ideas out of my mind with a thought partner rather than passively absorbing it.

Final Thoughts: Using Claude Honestly

I want to name the elephant in the room: intellectual integrity.

Claude makes it easy to look smart, but that’s not the point. The goal isn’t to make AI think for you; it’s to make you think with it.

This is the guiding principle for how I approach learning with AI. It should help me be smart, not just seem smart.

Everything I’ve described here is about using Claude to deepen your understanding, not to bypass the learning process. There’s a very clear line:

On the right side of the line are:

• Using Claude to understand concepts

• Getting help debugging your thinking

• Creating custom study materials

• Practicing with generated problems

• Building projects to test your understanding

Over the line are:

• Submitting Claude’s work as your own

• Using Claude to complete assignments you’re supposed to do by yourself

• Getting solutions to homework problems without attempting them

We’re the first generation of students to study with a ‘second brain’.

How we use that gift will shape not just what we know, but who we become.

Good prompting isn’t about getting the right answer. It’s about becoming the kind of thinker who knows the right questions to ask to extract insight as efficiently as possible.

That’s how to Claude.

#KeepThinking #Claude #Anthropic #AI #ClaudeAtStanford #Stanford #ClaudePartner

proofread with ❤️ by Jeremiah Nnadi + Adebayo A. + Olaoluwatomi Adesina (安天美), PMP + Rishabh Jha + Oluwasanmi Micheal-Giwa

We’re losing the beauty of vintage.

We no longer let things lose their shine but keep their purpose.

Consumerism compels us to replace, not repair.

Rather than letting things age with us—their cracks and dents mapping our shared years—we discard them.

We throw away our old toys, and with them, the parts of our lives buried in them.

That worn tee from a concert we can’t recall no longer sits quietly in the corner, waiting to jolt us mid-thought, reminding us of who we used to be, and how far we've come.

We’re losing a precious connection to our present selves by erasing the artifacts that once helped us converse with our past selves.

We’re losing the healing power of vintage.

What would it look like if we let our belongings age gracefully alongside us as quiet companions?

Not for hoarding, but for bearing witness to the unraveling of our wild, wondrous lives.

Read more

The past few weeks, I’ve watched a lot of docuseries.

Besides being some of Netflix’s best work, docuseries visualise my favourite genre of books—biographies.

To me, biographies are the ultimate genre of storytelling. They combine elements of fiction and non-fiction yet are neither at the same time. They’re the perfect literary paradox, and I’m obsessed with paradoxes wherever I can find them.

Another reason I love biographies is that they provide a sense of order to the chaos of life. They distil the seemingly random trials, tribulations, and triumphs of everyday life into a series of events that culminate in what seems like an ordered life. 

Biographies uncannily strengthen my belief in the invisible hand of fate just as much as in the power of self-directed ambition.

When you read…

Read more

One of the ways I can tell I’m in my peak depressed era is when my own company starts to terrify me.

As an introvert who moonlights as an ambivert, I recharge my social battery by pulling myself away from the crowd to sit with and converse with myself. It is in those moments of solitude that my most brilliant ideas and my most fearsome demons congregate in my mind.

I used to seek out solitude. I used to be deliberate about making time each day to chill with my oldest friend and foe––myself. In those moments, I could just sit and stare into space while recounting and reflecting on experiences from the day; or I could be ranting to myself about people and events that triggered me.

I was alone, but I wasn’t lonely. Not in the slightest. I could sit with myself for hours, sifting through my mental ca…

Read more

Read more

Subscribe now

Read more

Subscribe now

Read more

Subscribe now

Read more

Subscribe now

Read more

Subscribe now

This week, I’m going to write to you like I’m writing to my therapist. Think of this as an open journal entry that I hope you find helpful.

I’m at a point in my life where so much is changing. As is often the case with such transitional periods, there’s a lot of stress and anxiety.

I left my job at Facebook two weeks ago to start my master’s degree at Tsinghua University. As a 23-year-old who had gotten used to enjoying a generous paycheck at the end of each month, leaving my first full-time job to plunge into the uncertainty of the job market all over again wasn’t as easy a decision as I’d thought. 

On my last day at work, when it finally hit me that I was leaving the comfort of stability to jump headfirst into the unknown, I broke down. I cried during my final meeting with my manager, and pretty much the rest of th…

Read more

Subscribe now

Read more

Subscribe now

Read more