“A book must be the axe for the frozen sea within us.” - Franz Kafka

No one asked, but here are 20 books I would recommend. The list is inevitably shaped by my professional interests & the environment I grew up in, although plenty have nothing to do with either.

Some are easy reads, others took a fair bit of persistence - Tolstoy & Dostoevsky in particular needed a few attempts over the years. No ranking or set criteria, just books that have stuck with me for one reason or another.

1. Nineteen Eighty-Four by George Orwell

2. Crime and Punishment by Fyodor Dostoevsky

3. The Alchemy of Finance by George Soros

4. Baruch: My Own Story by Bernard Baruch

5. Iacocca: An Autobiography by Lee Iacocca

6. Principles by Ray Dalio

7. The Count of Monte Cristo by Alexandre Dumas

8. Extraordinary Popular Delusions and the Madness of Crowds by Charles Mackay

9. Only the Paranoid Survive: How to Exploit the Crisis Points That Challenge Every Company by Andrew S. Grove

10. Hedge Fund Market Wizards by Jack D. Schwager

11. Atlas Shrugged by Ayn Rand

12. The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers by Ben Horowitz

13. Hard to Be a God by Arkady & Boris Strugatsky

14. The Master and Margarita by Mikhail Bulgakov

15. Reminiscences of a Stock Operator by Edwin Lefèvre

16. The Great Game by John Steele Gordon

17. War and Peace by Leo Tolstoy

18. The Black Swan by Nassim Nicholas Taleb

19. The Power Law by Sebastian Mallaby

20. Lolita by Vladimir Nabokov

There is a peculiar loneliness to loss. Not the dramatic kind that arrives in a single liquidation or some spectacular blow-up, but the slow death by a thousand cuts. The sort where every morning you wake up believing tomorrow will be different, only to find yourself checking the same position, going through the same synthesis that led you to this place, and desperately attempting to convince yourself that patience is still the best course of action. Somewhere between the delusional conviction and inherent stubbornness, the line quietly disappears, and the loss grows larger before you even realise it.

Last year, I took a loss that still frustrates me to my core; simply thinking about it makes my blood boil. I know I am coming off extremely petty and bitter. Time has passed, markets have moved on and couldn’t care less about what I think, other opportunities have come and gone, yet it remains lodged somewhere in the back of my mind and deeply infuriates me. Not because it was life-ending, nor because I cannot recover financially, but because it challenged something far more valuable than capital or messing up a beautiful PnL streak. It challenged the belief I had in my own judgement, making me chase and operate on emotion; this is the path to death by a thousand cuts.

Losing money hurts; no surprise there. But it isn’t the capital. It is every decision that led me there. The conversations I should have had with myself. The position I definitely should have reduced. The other side I dismissed on the basis of my own bias. The trade I knew was not behaving adequately, yet I somehow found it within myself to convince myself that my line of thinking simply needed more time to manifest. And hey, I paid the price. I deserved it. I was the fish.

Perhaps that is why friends checking in almost feels worse than the loss itself. They mean well, but by the time they ask how you’re doing, you’ve already had the conversation with yourself a hundred times over. At that point, the ego has already absorbed the blow, or maybe it’s still processing it. You begin asking questions that have nothing to do with PnL.

Can I actually do this again?

Was the first outcome skill, luck, or some combination of both? Knowing what it took to build it the first time makes rebuilding feel immeasurably heavier, and the cut grows deeper every time you think about it.

Wins have an extraordinary ability to inflate the ego. You start believing your process is robust enough to survive almost anything (I am aware of how dumb this comes off). Loss has the opposite effect. It strips away certainty with remarkable efficiency. Even observations you were objectively correct about become difficult to trust due to the fact that you failed to act on them and execute them in time. There is something uniquely frustrating about being right in thought but wrong in execution. Hence, a valuable reminder: markets do not reward ideas; they reward action and dispassionate decision-making, even if it means closing the trade and thinking things through.

Prospect Theory offers a satisfying explanation. The asymmetry is remarkably intuitive once you experience it. It is simple to understand: we derive far less pleasure from gains than pain from equivalent losses. A winning trade quickly becomes your new baseline, which is a flawed standard to operate on. On the other hand, a losing trade stays with you for months, perhaps years, maybe serving as a memory of what we lost. It is difficult to explain to someone outside markets why one bad trade can outweigh twenty good ones emotionally, but anyone who has experienced it will understand immediately. I suppose this is the final phase of digesting the loss and processing what happened.

The emotional response rarely announces itself all at once. It arrives quietly. Losing trades gets you biting your nails, feeling sick to your stomach, pulling hair, breathing heavily, and slowly turning tilted. You begin trading from reflex and impulse rather than thought and structure. You become increasingly short-sighted, particularly if you remain structurally long-biased in a market perfectly willing to crucify optimism and punish any form of bullish expression. We can argue that each decision feels reasonable in isolation, yet collectively they compound into something much larger.

Looking back, some of the mistakes were blatantly obvious. Others resemble being scared of the monster under the bed. You spend months protecting yourself against unlikely outcomes while quietly ignoring the obvious risks sitting directly in front of you. There is another trap that remains surprisingly difficult to overcome: becoming hesitant to short something because it is already down or thinking you are too late to the party. Logical synthesis tells you it must be cheap, yet it is down for a reason. I might outright butcher the exact expression, but markets have an uncomfortable habit of remaining irrational far longer than one can remain solvent. Lessons in that.

Perhaps these are simply the classic thoughts you succumb to during a bear market, when prices remain generally compressed, and everything feels structurally broken. While I envision the depressed state and more aggressive exit from the space may have been rooted in exhaustion rather than conviction, it is remarkable how permanent temporary conditions can feel. Then, when the tide shifts, suddenly the same market that looked terminal only months earlier begins rewarding risk again.

What remains baffling to me is how reflexive it all is. Objectively, if people don’t care, they don’t care. Then, all of a sudden, they care all at once. These markets remain sanguine until they aren’t. Risk-on gets rewarded while somebody else inevitably plays the role of the sacrificial lamb (I definitely was). Timing this is challenging at best, which perhaps explains why so many participants convince themselves that conviction alone is enough, or even worse, they position themselves based on borrowed conviction.

Meanwhile, mouth breathers get rich and pontificate about trading and markets. There are a lot of fools peacocking these days. The old quote that “the main purpose of the stock market is to make fools of as many men as possible” feels increasingly relevant time and time again. Success and randomness often become indistinguishable over short periods. Heuristics become difficult to justify. If I were a gambling man, I would probably bet, and that’s precisely the problem.

The biggest lesson wasn’t about that specific trade. It was understanding the cost of tuition. You can’t be spending millions of dollars on lessons you could have learned before the trade. That’s outright stupid. There comes a point where self-punishment stops being productive and simply becomes another position you refuse to close, but in most cases it’s better to peel the plaster off early. The objective is not to forget, or even worse, pretend it never happened. The objective is to ensure you never pay for the same lesson twice.

The loss will sting, but not nearly as much as realising you were the fish. If you are a competitive and somewhat capable market participant, that’s the thought that truly lingers. Money can be earned back. Ego eventually heals, but you become more rational. Confidence slowly returns, even if your armour has a dent in it. But only if you remain emotionally intact throughout the process and don’t turn to self-destruction.

As prices perform well again, I hope to look back at this stream of thoughts in a different light, not as an obituary of my mistakes, but as a reminder of what markets extract from us when we become complacent and succumb to the trap of greed. The goal was never to avoid loss entirely. That is outright impossible; it’s part of the process. The goal is to have enough bullets in the gun to be able to take down the elephant when the time comes. To occasionally lose the battle, but survive to fight another day.

Vidura Niti says that wisdom belongs not to the man who never errs, but to the one who refuses to repeat the same mistake. Markets are unusually generous teachers. They allow us to fail repeatedly, blow up and make emotionally driven choices. They simply charge extraordinary tuition, and you can either deal with the pain or continue chasing and die a slow and painful death of a thousand cuts.

The lessons here are beyond the loss itself. I definitely feel bitter, and it is inherently evident in the tone of my writing. I deem it worth it if I am no longer the fish.

I drew inspiration to verbalise my interpretation of crypto’s demoralisation & structural weakness through Thiccy’s piece “dealing with loss” & Dougie’s piece “crypto is dead”.

“Losing money is what kills you. It’s not the actual loss. It’s the fact it messes up your psychology. You lose the bullets in your gun. Then the elephant walks by when your gun’s not loaded. In this game, you want to be there when the great trades come along.”
- Brett N. Steenbarger

It is an undeniable truth that most reap what they truly seek from markets.

Capitulation remains a habitual phenomenon. It serves as part of a natural cleanse we experience each cycle, as we begin to contemplate what follows.

Beginning to question crypto’s positioning as a category in its lifecycle -attempting to uncover the remaining foundations we have to show after years of inflated promises, manic belief, emotional rollercoasters & unrealised expectations.

Does crypto possess the underlying qualities to confidently belong on a pedestal alongside established & mature markets? Or are we eternally doomed to remain a capital black hole intrinsically driven by speculative value & gambling seeking an escape hatch from financial mediocrity?

I am well aware of my excessive cynicism, yet I only allude to our collective past failures in an attempt to articulate my vision for crypto’s continuity.

Participants are quick to draw analogies with the Dotcom crash, refer to the Gartner cycle, in an attempt to pinpoint our shortcomings & express potential optimism for the future.

Some reluctantly surrender to the urge of pointing fingers, looking for scapegoats. Falling to the extent of calling out accounts rocking NFTs on their Twitter (while conveniently turning a blind eye to NFTs in their portfolio) or disregarding their personal venture bags in L1s altogether when articulating their L1 doom thesis (a few characters immediately jump to mind, love Santiago, to be clear).

Everyone is entitled to an opinion, yet the extent of distress visible on the timeline, as well as in the orderbooks feel progressively worse, especially in light of the continued outperformance of other risk assets, while alt land continues looking ever gloomier.

The levels of coping are truly a miserable sight to witness.

• “I miss 2021”

• “Crypto is dead”

• “I wasted my last 8 years in crypto”

As dismissive & unempathetic as I may come off, operating under this frame is fundamentally flawed. It will be more about revanche than echoes - it always is.

The early discoverers of the land of magical internet coins have succumbed to the curse of being early. The right place at the right time enables a perfect setup & I sincerely hope those participants fully capitalised on that blessing, as the current iteration is significantly more nuanced & the game is now on hard mode.

“Skill issues” exist & we all cave in to over-index on past success, often failing to objectively reflect on the holistic picture the board presents.

The collapse of FTX presents a compelling case to revisit. The crash & burn of the exchange had a salutary effect on the market - a catharsis that allowed us to look forward & leave many of the foul actors, along with the sins of the last cycle, behind us. While it may have felt like we were roadkill, it objectively presented a unique prospect to seek asymmetry & figure out how the next fortune reversal might shape up.

Fast forward to today: the feedback loops of the market have accelerated exponentially, & the industry as a whole has shown signs of attempting to escape the status quo. This should be a moment of salvation - to leave the tricks & games of the past behind, rendering them largely obsolete.

I loved the parallel Scimitar Capital made in his piece:

“In Greek mythology, Sisyphus is condemned to spend eternity pushing a boulder up a hill, only to watch it roll back down the moment he reaches the top. There is something uniquely cruel about this punishment, something that cuts directly to the core of the human experience that drives many to suicide.”

Trading shares this quality. Unlike other professions, no checkpoints exist. A single faulty or misguided decision can doom a flourishing career built on hundreds of wins.

The single adjustment I would make is applying this parallel to both drawdowns & periods of stellar success.

We tend to underemphasise when you win big. When do you really step away? Do you at all? As the goal post progressively moves further away, accolades are hit that were once meant to be the end line.

A symbiosis of greed & addiction. Cognitive biases, amplified by volatility, driven by the addictive nature of information & price movement. Volatility - so deeply ossified within our space. These biases stem from Prospect Theory, gluing us to the table when, in reality, it is often best to step away.

But if you are reading this, the journey likely continues. Markets like these are the greatest testament to & measure of one’s true love for the game.

In my eyes, the most meaningful question to ask yourself is: Do you still have Edge?

It is clear we are in a structural bear market for alts. Even now, with valuations down exponentially, I am unable to shake the thought that we remain detached from reality.

We had ETFs, DATs, so what is the real reason to believe now?

There is no marginal buyer for the majority of coins, & the industry has transitioned to become more PvP than ever before.

I find it highly impractical to position purely on hope - that the money printer will drive miracles or that mean reversion will save us. These forces remain outside our control.

The market remains horrendously inefficient, & fundamentals remain largely questionable across crypto.

What is worth backing? The continuous existentialism persists.

Surely, as an industry, we are more than DeFi, stablecoins, perp DEXs, & potentially prediction markets?

The never-ending debates surrounding infrastructure vs app layer. The “revenue meta” is an effort to distinguish vaporware from substance, holding protocols to the standard of Hyperliquid.

Inevitably, ideas require funding. Jeff set an example - fully backing his own ideas, bootstrapping in the early days of the protocol - something that remains largely unaffordable to replicate.

I know contrasting views have been expressed on the topic of venture allocators, & I remain far too biased, yet venture backing is an integral part of the emergence of genuinely novel ideas & teams.

“People don’t know what they want until you show it to them”

⁃ Steve Jobs

There is a kernel of truth embedded in that quote that feels especially uncomfortable for crypto today. We have spent years oscillating between dogma & reactionary verbiage - mistaking loud narratives for progress & even consensus for conviction.

Too often, we ask users what they want, only to mirror back incremental abstractions of what already exists elsewhere - cheaper, faster, more leveraged versions of yesterday’s ideas. That has never been how new categories are born.

If crypto is to stand on parity with other asset classes, it will not be by winning marginal efficiency contests or endlessly recycling the same constellation of beliefs around infrastructure, incentives, & reflexive growth.

It will require novel primitives, new mental models, & a willingness to step beyond what already “works”. Not everything that matters can be product tested, revenue-mapped, or justified in a simple deck. Some ideas only reveal their inevitability in hindsight - & that takes time.

This is where the industry must grow up. Less reverence for past cycles, less nostalgia for 2021, & more intellectual truth about what is structurally broken - & what remains possible to rebuild.

The path forward is unlikely to look familiar, & it certainly won’t be obvious at the moment it matters most.

Yet, despite the noise, despite the unprecedented urge to capitulate, I remain hopeful. Markets have a way of humbling us right before they bless us.

Those who have seen it before understand this truth: participants will be crucified every time. Some will survive. Some will return.

If history rhymes at all, then perhaps what lies ahead is not an echo of what we lost, but a final, improbable last dance - reserved for those who stayed mentally intact, remained solvent, upheld the rigour as the game evolved, & kept their edge intact.

In the words of my friend Dougie,
Long Live Crypto.

A trader I hold in high regard recommended watching “Jiro Dreams of Sushi”. They drew an analogy between the granularity and continuous refinement of the process of the Japanese chef Jiro and that of the highest calibre of trader.

Initially, I took it to be somewhat of a stretch (to say the least). However, as I got on a 14h flight, I thought, given that I did not have many better things to do at that moment, I would give the documentary a watch and see if I could draw any major parallels between the 2 crafts.

Jiro is a chef in Japan. He is 100 years old and is the owner of Sukiyabashi Jiro. Jiro is unique; he is obsessive, truly harsh on himself and stubborn in his ways.

A good trader, similar to that of Jiro, hones in on his craft. This obsession for simplicity in the process with defined steps is the magic of a 3 Michelin star chef, and a trader carrying many wins and accolades under his belt, from my perspective, is no different.

In the documentary, it is outlined that the attributes of a great chef are:

1. Take work very seriously and perform at the highest level

2. Aspire to improve one’s skills

3. Maintain cleanness

4. Impatient character (they are often better leaders than collaborators)

5. Unprecedented passion for their craft

All of those apply to trading.

In the story of Jiro, his edge lies in the fact that he finds it essential to double-click on every single detail.

In the restaurant, they would age a small tuna for about 3 days and a larger tuna for up to 10 days. Everything is being done ahead of time, similar to the research and work it takes before any trade ever takes place.

What others might not notice, Jiro will spot simply from having done this for such an extensive period of time. Once again, if we draw a parallel, in a trade, if you are not meticulous in your ways, you will be more prone to losses and succumb to growing sloppy over time - discipline is essential.

His routine is simple, yet timeless. He has taken the same steps for over 50 years. The only occasions during which he takes time off and has days when he is away from the restaurant are national holidays and funerals.

“The way of the shokunin is to repeat the same thing every day”

This is rare, and I find it to be beautiful. In my eyes, this is exactly what it takes to stand out from the herd as a trader.

In France, the 1st Michelin guide was first published in 1900.

Michelin inspectors look for:

1. Quality

2. Originality

3. Consistency

Again, I know I am beginning to sound like a broken record here, but all of those in my eyes qualify as traits of a good trader.

For reference, a perfect 3-star Michelin rating means that it is worth making a trip to that country just to eat at that restaurant. You can probably understand that when Jiro’s restaurant received 3 Michelin stars in 2008, most were rather astonished.

Think about it.

There are only 10 seats in his restaurant, with the restroom being outside of the premises. The average duration of a meal is approximately 15-25min, with no appetisers, just sushi. We can definitely say that there is no other 3-star rated restaurants anything like this in the world, and it would be an understatement.

“No matter how many times you eat at Jiro’s, the sushi there is incredible” - Michelin

Right now, it might just seem like everyone is obsessed with Jiro, almost becoming deluded. Surely his food cannot always be a 10/10 every time? He is inherently dependent on external factors; what if they let him down? i.e bad tuna or shrimp? Before WW2, they used to cook fatty tuna (o-toro); the taste of the fatty tuna is simple & predictable, although the flavours of the leaner cuts (akami) are significantly more subtle and sophisticated.

Each tuna has its own unique taste, yet it is the leaner cut that carries the essence of the flavour.

“We can’t buy just any tuna” - Jiro

and a good trader cannot afford the complacency of bidding any random asset ;)

Now let’s dive into the supply of tuna. The tuna vendor is a priori the tuna expert; the same applies to the shrimp vendor, who is a specialist in his field. Each one of the vendors is a specialist in their niche and has a unique edge that Jiro and his crew do not have. It is a de facto that specialists, in this case, the vendors, are significantly more knowledgeable.

If we are honest, they are the ones who can disrupt the flow for Jiro and his culinary magic. To avoid such occurrences, the team made a conscious decision to build a relationship of trust with them.

Trust is everything. In markets, those same trust assumptions apply.

“I would make sushi in my dreams. I would jump out of bed at night with ideas” - Jiro

There is always a yearning to achieve more with Jiro. I find it honestly fascinating: the level of obsession and dedication to the craftsmanship.

1. Always try to look ahead and above yourself

2. Always try to improve yourself

3. Always strive to elevate your craft

Those are the lessons Jiro taught his sons. There is even an undeniable level of joy when a master calls the apprentice a “shokunin”; it is what they strive for over many years.

The way of the shokunin is powerful in its own right, and I believe we have a lot to learn from it & carry over many of the teachings to our own pursuits.

“Perfection is achieved not when there is nothing more to add, but when there is nothing left to take away.”

Lately, I have found myself overly cautious. Markets like this have a unique way of turning even the highest-conviction bull into a perpetual cynic. Inherently, most participants joined with an expectation of asymmetry, the premise of outsized returns, yet somehow instead found themselves stuck in recursive cycles of volatility and, to a larger extent, disappointment. Increasingly more, I hear the refrain that the market is “boring”. This really has a way of throwing me off.

Is it truly boring as so many claim?

Or has it simply been unrewarding for your bags, and we stopped having everything experience a parabolic moment and do silly numbers? 10-20x? (We did have Aster, Zcash, Plasma and other tokens melt faces in a similar fashion, fyi).

It is rather ironic that the more time I spend around markets, the more I view them through the same lens as poker. The top percentile hone in on optimising the process itself vs fetishising the result and outcomes like many of us succumb to in the space. In markets, this becomes exponentially harder; we largely operate with incomplete information, blurred and often delayed feedback loops, as well as path-dependent outcomes.

It is with a heavy heart that I say the addiction to volatility in my mind has deeply ossified into our culture and way of being.

We are in a place where we conflate action with progress, having exposure to developed conviction, randomness and honestly blatant luck, with a particular skillset. Inevitably misaligning timeframes and expressions.

For the most part, it is worth asking: Are you a trader? A persistent bag holder? Or just a hopeful one slightly high off copium?

The human mind is genuinely rather remarkable in its capacity to rationalise and then proceed to stomach things (personally guilty of periodically outsized copium) and honestly turns into the most efficient liquidity sink of all. I think operating under the assumption that most reap what they truly seek from markets is not a bad frame of thought, per se. Some seek validation or adrenaline or a sense of belonging, and I think for all of us universally, if you are playing the game, you seek capital. Capital preservation, let alone wealth creation, even a semblance of it, demands restraint, a kind of intellectual sobriety.

In the end, making money as an idea is awesome; it is mouth-filling. Objectively, who would not want to? Although in practice it is, in most cases, rather a tedious and boring process that requires the highest level of pedigree and, as with everything worthwhile, time. In my eyes, it is largely dependent on one’s discipline, patience and overall technical acumen.

Same as in poker, where the game will reward those who can detach from trepidation & recency bias, those who are here for the love of the game over the individual outcome and will continue no matter the weather, the time of day, nor the hand they are dealt.

I have had this line stuck in my head, and I think it is only appropriate to close out this spree of thought with it.

“The two hardest tests on the spiritual road are the patience to wait for the right moment & the courage not to be disappointed with what we encounter.”

I would like to articulate that this is not an attempt to over-intellectualise gamblers putting capital behind a new format of betting, but rather diving into the core of what Sperm Racing is looking to achieve, the problematic side of male fertility at the moment and the result of 16- and 17-year-old founders getting together to disrupt a vertical with one of the most unique, crazy and creative ways.

Beneath the meme-worthy absurdity, there is a crisis unfolding. Male fertility rates continue to tank, and it does not seem to be an area getting much light shed on it. According to Human Reproduction Update, global sperm counts have dropped by 59% between 1973 and 2011. More recent studies suggest the rate of decline has accelerated to >2.6% per year since 2000. At this trajectory, the average man could reach sperm counts below the threshold for fertility by 2045. This is a societal problem unfolding.

The causes are manifold, from endocrine-disrupting chemicals, processed food, sedentary lifestyles, chronic stress, nicotine use (guilty), rising obesity rates, to excessive screen time. Male testosterone levels have declined in lockstep, dropping nearly 1% per year for the last four decades. A study from the Journal of Clinical Endocrinology and Metabolism showed that a 60-year-old man in 2004 had testosterone levels 17% lower than a 60-year-old in 1987. Rather crazy to think about.

As a result, it is no surprise to see the rise of TRT clinics, once a medical solution for hypogonadism, now repackaged as a lifestyle upgrade. Joe Rogan, Dana White, Jeff Bezos, and many more have actively spoken out about their TRT usage. The narrative has shifted from pathology to optimisation. To think that society turned default into an intervention. In a way, “synthetic” testosterone is the answer not to disease, but to normal decline.

Meanwhile, IVF is no longer niche; it has turned pretty mainstream. The global IVF market is expected to surpass $36b by 2028 (what a great business), driven by delayed childbearing, declining fertility, and more same-sex couples and single parents opting for assisted reproduction. In Denmark, >10% of babies are born through IVF. In Israel, it is 1/9. The average IVF cycle costs over $15k in the US, and success rates still hover below 30% per attempt. Fertility is rapidly expanding to become a luxury good, subject to economic inequality.

And with that, the ghost of eugenics looms. The more we optimise, select, enhance - whether through embryo screening or hormone engineering- the closer we edge toward a society that treats reproduction not as a miracle, but a market. Sperm, like talent or intelligence, becomes selectable. Customisable and maybe even ranked. One of the most interesting companies I looked at recently does exactly that.

With all this said, enter Sperm Racing.

You might think that it sounds like a joke, but lowkey it is. But it is also not. The structure is real: microfluidic devices mimicking the female reproductive tract, chemical signals guiding sperm cells to a finish line, and optical sensors verifying the winner. It’s part spectacle, part science experiment. A blend of biology and sports.

And like any good spectacle, it mirrors sport. Think back to April 9, 2005. Forrest Griffin vs. Stephan Bonnar. UFC 1 Finale. That fight did not just entertain; it highly legitimised the UFC. It turned a niche, borderline illegal bloodsport into a multi-billion-dollar global league. Not through coordination or planning, but rather through virality.

Sperm Racing is made from the same cloth. It is wild, unpredictable, visceral, to say the least. Which also makes it perfect ground for betting.

Prediction markets are built on asymmetric information. But with sperm, performance is opaque. There is no footage of a sperm camp, no insights into last week’s motility drills. You can know the biometrics: age, fitness, testosterone levels - but it does not necessarily translate cleanly into odds. In the case of Noah Boat’s and Jimmy Zhang’s race, Jimmy is 2 years older, Noah is in arguably better physical shape, yet Jimmy won. This uncertainty creates real markets. You are not betting on skill; you are betting on biology, and one can argue on chaos. Not your typical casino; it is phenotypic probability meets internet and meme culture.

In a way, we can argue that is why it works. Virality is a key step, as it gets the foot through the door. More importantly, data comes in next. Each race generates metadata on sperm motility, morphology, and health - creating the first decentralised fertility dataset ever built in public. And over time, patterns emerge. An immense opportunity that, in a way, is being overlooked.

Age vs. speed. Diet vs. quality. Environment vs. morphology. And more.

Not something that might stand out initially, but it’s a real opportunity. I guess you could say what started as entertainment becomes infrastructure. What begins as a race ends as a research initiative. Sperm Racing is crazy, but that is kind of the point. Because fertility is not just science. It is culture. And sometimes, you have to laugh before you learn - being in crypto, many of us know that too well.

By no means is it a whitepaper on reproductive biology. It is exactly what it sounds like: sperm races. But underneath the wildness is a system that has been engineered with precision.

The race begins with a microfluidic device, a microscopic racetrack that simulates the female reproductive tract. Two sperm samples are loaded into identical channels. At the far end of each track is a chemical attractant - from my understanding, it is typically progesterone, the hormone naturally released by an egg to guide sperm, acting as the finish-line beacon.

To ensure a fair start, both sperm samples are held in place behind a barrier. At go-time, the barriers are dropped simultaneously. The sperm, now free to move, will swim towards the finish line guided by the chemical gradients and fluid dynamics of the racetrack. Everything, from temperature to fluid viscosity, is controlled to mimic a natural fertility environment.

All tracked in real-time. High-resolution microscopes paired with optical sensors and imaging software monitor movement, speed, and pathing. The race is broadcast with live commentary, stats overlays, and even slow-motion replays.

Once a winner is determined - the sperm that crosses the finish line first - the results are logged. The sperm samples are either ethically disposed of or, if previously agreed upon, donated to a sperm bank. It is equal parts science fair, performance art, and new-age fertility analytics.

In a world defined by spectacle, maybe that is exactly what we need. Definitely looking forward to the next race and the show the guys have set for us.

Fertility is in a bad place. I’m hoping that Sperm Racing can be a spark in people’s minds to begin actioning the problem at hand.

An agent can reason, browse the internet, write code and execute increasingly complex workflows. Yet the moment it needs to pay someone, access capital, or outsource something it cannot do itself, that autonomy begins to break down rather drastically. The existing financial system assumes that a person or legal entity ultimately sits behind every transaction, and understandably so due to the inherent trust assumptions. AI agents do not fit particularly neatly into that architecture.

This creates an interesting bottleneck. If agents are eventually going to operate businesses, manage workflows and participate independently in the economy, being able to think and act is only part of the equation. They also need the ability to transact in a structured and safe manner.

Payman is building the infrastructure for them to do exactly that.

The easiest way to think about Payman is as infrastructure allowing AI agents to transact with humans. An agent might be capable of completing 95% of a workflow itself, but eventually encounter something it cannot do. Perhaps it needs someone to physically inspect something, verify a piece of information, label data, make a phone call or perform some other task requiring human judgement. Today, that usually means the autonomous part of the workflow stops.

Payman instead gives the agent access to both capital and human labour. Through its API, an AI system can create a task, allocate money towards completing it and source someone capable of doing the work. Once completed, the task can be verified and the worker paid. The important distinction here is that the human does not necessarily initiate any part of this process. The agent does.

That sounds like a relatively small change. I think it is considerably more important than it might initially appear. If agents are eventually going to operate businesses, manage workflows or act on our behalf, they cannot simply be good at thinking; they need to be capable of acting. Money is part of that.

There is an understandable tendency to frame AI as a replacement for human labour. I am not convinced that is the most useful way to think about the transition, at least in the near term. AI is exceptionally good at some things and remarkably bad at others. Humans have the inverse problem. The more interesting system is therefore probably not one in which either operates independently, but one in which agents can dynamically pull humans into a workflow whenever necessary.

Payman effectively turns this into infrastructure. Human workers can make their skills and availability accessible to agents. An agent can then determine what it needs, post the task and select someone capable of completing it. This creates a peculiar inversion of the labour marketplace. Historically, humans have hired humans. More recently, humans have begun hiring AI. Payman is betting that AI will increasingly hire humans, as futuristic and somewhat counterintuitive as that may come off.

The marketplace therefore becomes less about people searching for jobs and more about autonomous software allocating work. That is a fairly interesting idea to me.

The second part of the problem is payments. Giving an AI the ability to decide that money should be spent is one thing. Giving it the infrastructure to actually move that money is another. Payman allows developers to fund an agent and programmatically give it spending capacity. The platform provides infrastructure around funding, task allocation and payments, with transactions able to move through both traditional banking networks and cryptocurrencies.

This matters because the payment itself should ultimately be almost invisible. An agent should not particularly care whether the person completing a task wants dollars in a bank account or crypto in a wallet. It should care that the task costs $50, that the person can complete it, and that payment can be made once the work is fully verified. The underlying rail becomes an implementation detail more than anything else.

This is also where crypto becomes considerably more interesting than another speculative use case. Autonomous software is inherently internet-native. Giving software access to internet-native money is fairly intuitive. But Payman does not need to make a binary bet between crypto and traditional finance. Supporting both allows the agent to operate across whichever financial infrastructure the human on the other side already uses.

There is another problem hiding inside all of this. If an agent hires a human, how does it know the work was actually done?

Payman’s initial architecture uses human verification agents to review completed tasks before payment is released. Combined with reputation data around workers, this creates a feedback loop around reliability and performance. The product shown at the time already demonstrated this basic workflow: an agent creates a task through the API, humans complete it, the work is verified, and payment follows.

It is admittedly somewhat recursive. AI hires a human to perform a task, another human may verify the task, and software coordinates the entire interaction. Over time, presumably more of the verification layer itself can be automated. But that is not particularly important for the initial thesis. The point is to allow agents to complete workflows today, including the portions where AI remains insufficient.

Waiting for models to become capable of doing everything is one solution. Giving them the ability to ask humans for help is another. The latter seems considerably more practical.

The timing is what makes the idea particularly interesting. AI agents are moving from something largely discussed conceptually towards products people are actually trying to build. The constraint increasingly shifts away from whether an LLM can generate an answer towards whether an agent can reliably take actions in the real world. Payments sit directly in that transition.

Payman had already attracted significant early interest, with the original thesis noting a waitlist of more than 10k users. The important part is less the absolute number than what it suggests: developers were already beginning to encounter the problem Payman was designed to solve.

There are obviously several things that need to go right. Agents need to become useful enough that people are comfortable giving them meaningful autonomy. Developers need to want financial infrastructure built specifically around agents rather than simply adapting existing payment APIs. There needs to be enough demand for human intervention to justify a marketplace. And perhaps most importantly, giving autonomous software access to money introduces an entirely new set of security, fraud and regulatory problems.

None of these is trivial.

But the underlying bet does not require believing that AI replaces everyone. Almost the opposite. It requires believing that AI becomes sufficiently capable to coordinate meaningful economic activity while remaining imperfect enough to occasionally need us.

That seems like a reasonably defensible place to start.

If agents become autonomous economic actors, they will need access to the same primitive that coordinates most human economic activity: money. And if they cannot complete everything themselves, they will need some mechanism for converting that money into human effort.

Payman is attempting to build the layer between the two. Perhaps the future of work is not simply humans working for companies or AI working for humans. Perhaps, increasingly, we work for the machines too.

The idea is hardly new, but the 2024 U.S. election has pushed it much closer to the mainstream. Polymarket became one of the more visible sources of election probabilities during the campaign, allowing people to trade everything from the eventual winner to individual states and political appointments. Its main presidential market alone ultimately attracted more than $3.5 billion in volume. The interesting thing was not simply that people wanted to bet on an election; people have been doing that for centuries. It was that a prediction market increasingly became something people looked at to understand the election itself.

That raises a more interesting question. Are prediction markets simply another form of speculation, or are markets actually a better way of finding out what people believe?

The Price of Being Right

Prediction markets are based on a fairly intuitive idea. Suppose a contract pays $1 if Donald Trump wins an election and nothing if he loses. If that contract trades at $0.60, the market is roughly expressing a 60% probability of him winning. That probability is not produced by a pollster or a model. It emerges from people buying and selling the contract.

There is a fairly important difference between asking someone what they think will happen and asking them to risk money on the answer. Opinions are largely cheap, and there are plenty of fools peacocking these days. Someone can tell a pollster, journalist or their friends almost anything without bearing much cost for being wrong. Markets introduce a direct and rather dangerous consequence. They do not necessarily discover truth because traders are smarter than everyone else, but they create an environment where people are financially incentivised to find information, update their beliefs and act when they think everyone else is wrong.

If you believe an event has a 70% probability of occurring while the market prices it at 50%, there is an obvious trade. If you are right, you make money; if you are wrong, you lose it. The mechanism turns disagreement into price discovery, and there is something quite powerful about that distinction between stated belief and revealed belief. I can tell you I think something is going to happen. The more interesting question is what price I am willing to pay for being right, the burden of truth is costly.

Markets as Information Aggregators

This idea goes back considerably further than crypto. Friedrich Hayek’s The Use of Knowledge in Society argued that much of the information required to make economic decisions is dispersed between individuals and can never be completely known by a central authority. Prices provide a way of compressing that dispersed knowledge into something everybody can observe. Prediction markets apply essentially the same logic to events rather than goods.

One trader may understand polling methodology. Another may know a particular state extremely well. Someone else might understand voter registration data, campaign finance, demographics or betting markets. None needs to possess the entire picture. They simply trade on the piece they believe the market is getting wrong, and the resulting price incorporates those competing views.

There is some historical evidence that this can work surprisingly well. The Iowa Electronic Markets began running real-money election markets in 1988 and became one of the longest-running experiments in prediction markets. Research comparing its forecasts with 964 national polls across five U.S. presidential elections found that the market was closer to the eventual two-party vote share than the corresponding poll 74% of the time, with its relative advantage particularly pronounced further from Election Day.

That does not mean markets are always better than polls. They measure different things and prediction markets themselves can be wrong. But it suggests there is something valuable about forcing information through a market rather than simply asking people what they think.

Polymarket

Polymarket takes that fairly old idea and packages it inside crypto infrastructure. Markets are generally structured around discrete outcomes, where a winning share ultimately pays $1 while a losing share becomes worthless, allowing prices between zero and one to function as market-implied probabilities. Traders use USDC and trade against one another rather than against Polymarket acting as the house. The result sits somewhere between an exchange, a betting market and an information product.

The 2024 election demonstrated how compelling that can become when the underlying event is large enough. Polymarket’s presidential winner market ultimately recorded more than $3.5 billion of volume, while countless related political markets traded alongside it. The obvious interpretation is that there is enormous demand for political betting. I think the more interesting interpretation is that there may also be demand for a continuously updating market price on uncertainty.

A poll is a snapshot, whereas a market is constantly adjusting. Every new poll, speech, endorsement, piece of economic data or breaking news can immediately be expressed through a trade. Participants do not have to wait for a forecasting model to be updated; the price simply moves. That begins to make a prediction market useful even to people who never place a bet.

There is an obvious temptation to take this argument too far. A market price is not truth; it is simply the price at which buyers and sellers currently agree to trade. For that price to contain useful information, the market needs enough liquidity and enough disagreement. Thin markets can be pushed around by relatively small amounts of capital, and participants with large positions can temporarily move probabilities without possessing better information.

This became a fairly visible debate during the election as several large accounts accumulated substantial Trump positions. The existence of a whale does not automatically make the resulting price wrong; someone willing to put millions behind a view may possess genuine conviction or information. But it complicates the idea that every market price represents some perfectly democratic wisdom of the crowd. Markets weigh opinions by capital, not by person.

That is both their strength and their weakness. Someone who believes they possess superior information can express that conviction more strongly than somebody making an uninformed guess. At the same time, someone with enough money can have an outsized effect on a market, particularly where liquidity is limited. The useful question is therefore not whether prediction markets can be manipulated. Every market can be manipulated at some scale. It is whether there is enough liquidity and enough financially motivated opposition for incorrect prices to become attractive trades. A manipulated price should, in theory, create an opportunity for somebody else. The difficult bit is making sure somebody is there to take it.

But Who Decides What Happened?

There is another problem which becomes increasingly important as prediction markets expand beyond simple questions. The blockchain can verify that a transaction happened on-chain. It cannot independently know who won an election, whether a politician resigned, how many times someone posted on X or whether Taylor Swift announced an engagement. Someone has to tell it.

Polymarket uses UMA’s Optimistic Oracle to resolve markets. Rather than relying on a single centralised source that simply declares an answer, an outcome can be proposed and challenged, with disputed resolutions moving through UMA’s dispute process. It is a clever solution, but it does not eliminate the underlying oracle problem. Prediction markets ultimately need to convert messy real-world events into discrete outcomes, and while sometimes that is trivial, in other cases language, timing and interpretation matter considerably. The more unusual the markets become, the more important the wording of the question and its resolution criteria become.

A prediction market can price uncertainty remarkably well and still fail if nobody can agree on what actually happened.

Post the Election

The larger question for Polymarket now is probably not whether the election was successful. It clearly was. It is what happens next. Prediction markets have historically struggled with a fairly obvious problem: the world does not produce presidential elections every week. Attention concentrates around major political, sporting and geopolitical events and liquidity tends to follow that attention. The 2024 election may therefore have been the perfect product for Polymarket rather than proof that every future event deserves a market.

The scale of the election effect becomes particularly obvious when looking at prediction-market volume over time. Activity had been gradually growing, but the acceleration into the election was in a completely different order of magnitude.

The original question therefore remains the important one: can prediction markets sustain engagement outside peak events like elections, or are they inherently cyclical, with activity following the news cycle? Historically, these markets have seen activity concentrate around elections and major geopolitical events before declining afterwards. Non-election volumes have been growing, which is encouraging, but at this stage they remain well below the levels generated by major political markets.

That does not necessarily mean every market needs to look like the presidential election. There is already activity around crypto, sports, economics, geopolitics and increasingly stranger cultural questions. Some of these markets are useful forecasts; others are closer to entertainment. I am not convinced that distinction matters as much as it initially seems. Financial markets have always combined information, speculation and entertainment to some degree. People trade because they want to make money, because they believe they know something, because they want to hedge an exposure, or simply because having a position makes watching an event more interesting. Prediction markets may ultimately work for many of the same reasons.

The more interesting possibility is that prediction markets eventually become infrastructure rather than destinations. A probability is useful information. Companies could use markets to estimate whether products will ship on time, investors could trade probabilities around regulatory approvals or corporate events, insurers could use event markets to price particular risks, researchers could create markets around scientific outcomes, while institutions could use them as another signal alongside surveys and models. The financial product and the information product begin to converge.

There is precedent for this idea inside companies. Internal prediction markets have been experimented with as a way of aggregating employee knowledge around things like project completion and sales forecasts. The underlying logic is the same: the relevant information already exists somewhere across the organisation, but nobody necessarily knows who has it. A market does not need to know who has the information. It simply needs to make being right sufficiently valuable.

Humans, For Now

There is another participant worth thinking about: AI agents. Most prediction markets today are still predominantly markets between humans, but that may not remain true for very long. An AI agent capable of continuously ingesting news, polling data, financial information and social media could theoretically maintain probabilities across thousands of markets simultaneously. Unlike a human trader, it does not need to sleep, become bored or decide that a small market on some obscure political appointment is not worth watching.

Give those agents capital, and they become market participants. That could create an unusual form of liquidity. Humans generate information and disagree about what it means; machines continuously trade those disagreements into prices. An agent might notice a new poll, compare it against historical polling errors, observe a move in related markets and adjust its positions before a human participant has even opened the article.

The result would not necessarily be omniscient markets. Models can share the same biases, consume the same bad information and become reflexive in exactly the same way human traders do. But prediction markets seem particularly well suited to them. The entire product is already machine-readable: a question, a set of outcomes, a probability and a payoff. This was already part of the thesis in November 2024: AI agents could eventually provide prediction markets with a new source of liquidity by making rapid probabilistic trades against real-time information.

Predicting Tomorrow

Prediction markets sit at an unusual intersection of speculation and truth-seeking. They work precisely because they do not ask participants to be impartial. Everyone is allowed to have an agenda, a bias or a view. The market simply attaches a cost to being wrong and a reward to identifying where everyone else might be wrong.

That does not make them infallible. Liquidity matters. Market structure matters. Whales matter. Resolution mechanisms matter. Regulation matters. A probability displayed on a screen should not suddenly be treated as objective truth simply because money sits behind it. But there is something compelling about the underlying mechanism.

We spend an enormous amount of time asking people what they think will happen. Pollsters ask voters, analysts publish price targets, economists publish forecasts, journalists interview experts, and investors write investment memos. Most of those opinions disappear with remarkably little consequence for the person making them. Prediction markets ask a slightly different question: what do you believe will happen, and at what price are you willing to be wrong?

The 2024 election has given us perhaps the clearest demonstration yet of what happens when enough people are willing to answer. Whether that activity survives the election cycle remains to be seen. If it does, prediction markets may become something considerably more interesting than a place to bet on politics. They could become a continuously updating layer for measuring uncertainty itself.

For now, the experiment is still early. But there are worse ways of trying to predict tomorrow than giving everyone a reason to tell the truth.

TON approaches the problem from almost the opposite direction.

Rather than building a network and subsequently searching for distribution, The Open Network sits alongside one of the largest existing distribution channels on the internet. Telegram has roughly 950 million monthly active users, an enormous portion of whom already use the application for communication, communities and, particularly within crypto, discovery. The interesting question is what happens when financial applications can be inserted directly into that existing behaviour.

Over the past several months, we have started to get an answer.

Notcoin attracted tens of millions of users through a remarkably simple Telegram Mini App. Hamster Kombat claims more than 300 million players. Telegram has introduced advertising revenue sharing paid in TON, USDT has launched natively on the network, and wallet infrastructure increasingly allows users to interact with crypto without leaving an interface they already understand.

Some of these numbers are difficult to reconcile with what we normally consider meaningful crypto adoption. Tapping a screen in anticipation of an airdrop is obviously not equivalent to becoming a recurring on-chain user. But dismissing the entire phenomenon for that reason probably misses what TON has demonstrated.

It has made participating in crypto unusually easy.

Whether it can make that participation worth returning for is a much harder question.

A rather complicated family tree

TON’s relationship with Telegram is both the reason the network is interesting and the reason it needs to be understood carefully.

Telegram originally developed the Telegram Open Network after raising approximately $1.7 billion from investors in 2018 through the sale of rights to its proposed Gram tokens. The ambition was considerably broader than simply adding payments to Telegram: TON was conceived as a high-performance blockchain architecture capable of supporting applications and payments at the scale of Telegram’s existing network.

That plan ran directly into the SEC.

In 2019, the Commission sued Telegram over the Gram offering, arguing that the planned distribution constituted an unregistered securities offering. A federal court subsequently prevented the distribution, and in 2020 Telegram agreed to return more than $1.2 billion to investors and pay an $18.5 million civil penalty. Telegram formally stepped away from the project.

The story did not end there.

Telegram had released TON’s code as open source. Independent developers continued working on it, eventually producing the network we now know as The Open Network, with Toncoin rather than Gram as its native asset. The TON Foundation emerged to support its development.

This distinction matters. TON is not simply “Telegram’s blockchain”, even though that description is tempting. Telegram and the TON Foundation are formally separate organisations. At the same time, pretending there is no meaningful relationship between the two is equally unhelpful. Telegram endorsed TON as its Web3 infrastructure in 2023, and TON-based services have increasingly been integrated into the application.

The relationship is therefore somewhat unusual: legally and organisationally distinct, but economically and strategically increasingly intertwined.

That tension sits at the centre of the thesis.

Distribution, distribution, distribution

Most Layer 1s have historically competed on some combination of throughput, fees, liquidity, developers and incentives. These things matter, but they have also become increasingly commoditised.

TON has something much harder to manufacture: distribution.

Telegram already has the users. More importantly, it already has the social graph through which applications can spread.

A Telegram Mini App can be discovered through a group chat, launched inside Telegram and shared with another user without asking either person to download a new application. Bots can communicate directly with users. Payments and rewards can sit alongside existing social behaviour. The application doesn’t need to convince the user to leave Telegram before it can begin onboarding them.

This substantially changes the funnel.

The conventional crypto onboarding flow might involve discovering an application, downloading a wallet, creating and securing a seed phrase, funding the wallet through an exchange or bridge, switching to the correct network, and finally interacting with the product.

Every additional step is an opportunity to lose the user.

Telegram allows much of the application experience to happen before the user necessarily understands that there is a blockchain underneath it.

That might sound almost trivial. I don’t think it is.

The best consumer infrastructure tends to become invisible. Most people using Spotify do not think about content-delivery networks. Someone sending money through Revolut does not need to understand the banking infrastructure settling the transaction. Crypto has often expected the opposite: users are exposed to chains, gas, bridges, wallets, signatures and addresses before receiving much utility in return.

TON’s more interesting proposition is that perhaps they shouldn’t be.

Notcoin and the tap-to-earn machine

The clearest demonstration so far has been Notcoin.

Notcoin began as an almost absurdly simple Telegram game: open a Mini App and tap a coin to accumulate points. There was little pretence of sophisticated gameplay. The attraction was the expectation that participation would eventually translate into an on-chain token.

It worked.

By the time NOT launched in May, the project reported more than 35 million participants, with the token subsequently listing on major exchanges. The important part was not whether tapping a digital coin constituted a good game. It was how quickly Telegram allowed an application with a financial incentive to spread through its social graph.

Hamster Kombat has subsequently taken the model considerably further. By July, Telegram’s Pavel Durov said the game had reached 239 million users in roughly three months; the project now claims more than 300 million players. The numbers are extraordinary, even allowing for the obvious questions around bots, multiple accounts and what exactly constitutes an active player.

What we know is that financial incentives combined with Telegram distribution are extremely effective at generating participation.

The model also has an obvious weakness. If the primary reason to use an application is the expectation of receiving a token, the application eventually has to answer what happens after the token arrives.

Crypto has seen this movie before.

Liquidity mining, play-to-earn and points programmes can manufacture impressive early metrics because users are being paid for behaviour. The difficult part is separating acquisition from retention. Incentives can introduce somebody to a product; they cannot indefinitely substitute for the product itself.

That makes the current tap-to-earn cycle simultaneously impressive and fairly difficult to underwrite.

Farming without knowing you are farming

There is nevertheless something genuinely novel here.

TON and Telegram have dramatically lowered the level of sophistication required to become an airdrop farmer.

Historically, farming a new network might require bridging assets, interacting with smart contracts, managing multiple wallets and understanding enough about crypto infrastructure not to accidentally lose your money. On Telegram, a user can participate in essentially the same speculative behaviour by opening a Mini App and pressing a button.

That sounds ridiculous because, to some extent, it is.

But reducing a complicated behaviour to a button is exactly what good consumer software tends to do.

The question is what happens after that abstraction has brought the user in. If somebody’s first interaction with crypto is earning points inside Hamster Kombat, can the same interface subsequently introduce them to payments, savings, commerce or other financial applications?

That transition is considerably more interesting than the clicker game itself.

USDT may matter more than the games

This is where I think stablecoins become important.

Tether launched USDT on TON in April 2024, placing the most widely used dollar stablecoin directly alongside Telegram’s distribution. The significance is not that Telegram users suddenly need another blockchain on which to hold USDT. There are already plenty.

It is that Telegram already functions as communication infrastructure for hundreds of millions of people, including communities where cross-border payments remain expensive or inconvenient.

Communication and payments have always sat unusually close together. WeChat demonstrated how naturally messaging can evolve into payments when the financial layer is embedded deeply enough into the social product. Telegram has the beginnings of a similar primitive, although obviously without anything approaching WeChat Pay’s adoption today.

A user sending USDT to somebody they already speak to on Telegram is conceptually much easier to understand than sending an ERC-20 to a hexadecimal address on a network they have never heard of.

This is the part of TON I find considerably easier to take seriously than tap-to-earn.

Games may be the acquisition channel. Stablecoins could be the utility.

Telegram as an economy

There is another piece of the ecosystem that receives less attention than the games: Telegram already contains things people are willing to pay for.

Fragment allows Telegram usernames and anonymous numbers to be bought and sold using TON. Telegram has also introduced advertising revenue sharing for channel owners, with payments made in Toncoin.

This matters because it gives TON something many crypto ecosystems lack: economic activity tied to a product that exists independently of the blockchain.

Telegram does not need somebody to invent a reason for a valuable username to exist. Valuable usernames already exist because Telegram has hundreds of millions of users.

It does not need to manufacture creators who want to monetise audiences. Those creators already operate Telegram channels.

Crypto can then become the settlement layer underneath behaviours that already exist rather than the reason those behaviours have to be invented in the first place.

That distinction is worth dwelling on.

A lot of crypto applications begin with the token and work backwards towards finding something for people to do with it. Telegram begins with communication, communities, advertising, digital identity and distribution. TON has the opportunity to financialise pieces of an economy that already exists.

Whether users actually want that financialisation remains to be seen.

The architecture is different for a reason

TON is also technically quite different from the EVM ecosystems most crypto developers are accustomed to.

The network uses an asynchronous architecture built around message passing between smart contracts. Contracts behave more like independent actors: they receive messages, update their own state and send messages onwards. Transactions that appear conceptually atomic on an EVM chain may therefore unfold as a sequence of messages across TON.

The architecture is related to TON’s broader approach to scalability. The network was designed around dynamic sharding, allowing work to be distributed across multiple shardchains as demand grows.

In theory, this architecture makes sense for what TON wants to become: infrastructure capable of servicing consumer applications operating at enormous scale.

In practice, complexity moves somewhere rather than disappearing.

Developers accustomed to Solidity face a different execution model and languages such as FunC, alongside different assumptions around asynchronous messaging and contract behaviour. The ecosystem’s tooling, libraries and infrastructure remain considerably less mature than Ethereum’s.

Distribution can attract users.

It does not automatically attract developers.

TON ultimately needs both.

The DOGS stress test

The end of August provided a fairly timely reminder of the gap between theoretical scalability and production infrastructure.

TON stopped producing blocks twice within roughly two days as the network dealt with extraordinary transaction load associated with the DOGS token distribution. One outage lasted around six hours; block production was disrupted the following day again as validators struggled with the load.

There is an irony here.

The network is explicitly designed around the possibility of bringing enormous consumer activity on-chain. Then one of the first genuinely enormous consumer distributions arrived and contributed to the network losing consensus.

That does not mean TON cannot scale. Early networks break, particularly under workloads they have not previously experienced. Ethereum itself has a long history of applications exposing infrastructure constraints, CryptoKitties being the obvious early example.

But it does make the recent outages particularly relevant.

TON’s advantage is distribution at a scale most blockchains can only imagine. Its infrastructure therefore has to survive workloads most blockchains never encounter.

The thing that makes TON interesting is also likely to be what tests it hardest.

Telegram without TON; TON without Telegram

There is another risk which I think deserves considerably more attention.

How much of TON’s value exists independently of Telegram?

This became particularly difficult to ignore following Pavel Durov’s arrest in France on 24 August. The details of the case are primarily about Telegram rather than TON, but the market’s reaction exposed how closely investors perceive the two.

Formally, Telegram and TON are separate.

Practically, much of the TON thesis depends on Telegram continuing to integrate, distribute and support TON-based applications.

That creates an unusual dependency for an independent Layer 1. Ethereum does not depend on a single consumer application for distribution. TON’s greatest competitive advantage is unusually concentrated.

The relationship cuts both ways.

If Telegram continues integrating TON, very few crypto networks have access to a comparable distribution channel. If that relationship weakens, through regulation, strategy or otherwise, TON becomes a considerably more conventional Layer 1 competing for developers, liquidity and applications against ecosystems with much deeper infrastructure.

The appropriate way to think about Telegram is therefore not simply as an advantage.

It is both TON’s moat and its dependency.

What are the users actually worth?

That leaves perhaps the hardest question.

TON can produce user numbers that make almost every other blockchain look irrelevant. But blockchain metrics are particularly easy to misunderstand when users are heavily incentivised.

A Telegram account is not necessarily a wallet. A Mini App user is not necessarily an on-chain user. An on-chain user is not necessarily economically valuable. And somebody tapping a hamster in anticipation of an airdrop should probably not be valued in the same way as somebody repeatedly sending USDT, paying for Telegram services or interacting with applications after incentives disappear.

The distinction between distribution and monetisation is therefore important.

TON has demonstrated distribution.

The next stage is proving what that distribution is worth.

I would pay much closer attention to stablecoin balances and transfers, recurring wallet activity, application retention after token distributions, developer growth and economic activity that does not require continuously issuing new incentives.

Those numbers tell us considerably more than registrations.

Distribution as infrastructure

There is an old line in startups that distribution matters as much as product. Crypto occasionally behaves as though this rule does not apply.

We have spent years building technically impressive infrastructure and then attempting to bribe users into discovering it.

TON is an interesting inversion of that model.

The distribution already exists.

Telegram has hundreds of millions of users, enormous network effects and a particularly strong presence within crypto itself. TON provides a financial and computational layer that can increasingly sit beneath that existing network. Mini Apps make applications discoverable through social interaction. Stablecoins make value transferable through the same interface. Fragment turns pieces of Telegram’s existing digital economy into on-chain assets.

The pieces fit together unusually well.

What remains unclear is whether they produce an enduring economy or simply an extraordinarily efficient machine for distributing tokens.

TON does not need to onboard a billion people into crypto.

If the thesis works, many of them may never need to know they are using it.

That, more than the clicker games, token incentives or headline wallet numbers, is what makes the experiment worth watching.

In November 2022, crypto exchange FTX filed for Chapter 11 bankruptcy protections as it attempted to assess the value of its remaining assets collapsed. The failed exchange was soon found to have an over $7b shortfall in assets due to bad loans made to sister market-making firm Alameda Research with customer deposits.

In the wake of 2022’s tumultuous events, epitomised by the downfall of FTX, the crypto community faced a stark reminder of the inherent risks in centralised exchanges. FTX’s collapse was not just a one-off scandal, but the continuation of a troubling trend of centralised exchange collapses. This incident underscored a fundamental flaw in the opaque nature of centralised exchanges: the over-reliance on trust-based custodial systems. Trust, which in this case ultimately ended in fraud and mismanagement.

The Rise of DEXs

The ethos of crypto has always been rooted in decentralisation and the elimination of trust-based systems. DEXs embody this principle by enabling permissionless peer-to-peer transactions without intermediaries, mitigating the risks associated with asset exchange. The trustless exchange of one crypto asset for another enables a new way of trading assets, enabling a new way to bootstrap liquidity for a market. The growth in DEX usage reflects a broader industry shift as users seek more control over their assets and a reduction in counterparty risk.

Over the past 4 years, we have seen DEXs rise from a large niche hobbyist market to one of the primary ways transactions are settled in the crypto space. The rise of Automated Market Maker DEXs, led primarily by Uniswap, radically simplified the DEX experience, making it easy for anyone to perform a swap in a few clicks. In 2023, the portion of DEX volume relative to CEX volume crossed over 20%, reaching an all-time high driven by on-chain trading of meme coins on Ethereum.

DEXs have grown considerably in recent years, and the market is considerably larger when looking at broader centralised exchange volumes. The opportunity for DEXs remains to eat into the CEX market, which at its peak in May 2021 rose above $4 trillion in volume and over $1.1 trillion in December 2023. While CEXs historically have had an advantage over DEXs due to superior liquidity and a smoother user experience, DEXs today have an opportunity to dramatically simplify this experience, swapping one asset for another wallet to wallet. For the space to reach its full potential, there is a need for permissionless swaps.

The Future for DEXs is Cross-Chain

DEX volume today remains largely exchanging assets that exist on the same chain; however, the dream for DEXs has always been true cross-chain swaps, allowing direct exchange of one asset for another between different blockchains. While bridges offer one solution to this problem, they do come with their own trust assumptions and security risks, which may not always be clear to the end user. Seamless cross-chain swaps of one asset for another remain the gold standard for lowering the barriers between blockchains. Crypto is rapidly evolving towards a multichain future, necessitating platforms that enable seamless cross-chain interactions.

Chainflip: The Cross-Chain Liquidity Network

Chainflip is a decentralised protocol designed to address the challenges of cross-chain swaps in crypto. It stands out by enabling seamless and secure exchanges of assets between diverse blockchains, directly from users’ wallets, bypassing the need to rely on centralised exchanges or bridging solutions. Chainflip leverages a dual-layer structure comprising an Accounting Layer and a Settlement Layer. The Accounting Layer, operating on a proof-of-stake consensus, maintains the protocol’s state, including tracking deposits, withdrawals, and swaps.

Chainflip simplifies the user experience of cross-chain swaps by eliminating the complexities associated with wrapped tokens and specialised wallets. The protocol achieves this through a robust network of validators and the use of Threshold Signature Schemes, minimising reliance on centralised entities and single points of failure.

Why is Chainflip better than existing approaches to cross-chain swaps?

Unlike some existing IBC-reliant approaches, Chainflip offers true interoperability between different blockchains without requiring any modifications to the existing chain infrastructure. Chainflip achieves this by facilitating swaps directly from wallet to wallet without using wrapped tokens or specific blockchain integrations, which can limit their scope and increase complexity. This approach also dramatically improves Chainflip’s user experience, with users performing swaps directly from their wallets, without transferring assets to a centralised exchange or navigating complex bridging protocols.

The Settlement Layer, known as the Chainflip Engine, manages the actual transfer of assets across different blockchains. Central to its operation are validators, collectively forming the Authority Set, chosen through auctions based on their stake in the native token, FLIP. These validators are crucial for both layers, holding private keys for consensus participation and Multi-Party Computation (MPC). The protocol utilises vaults for liquidity storage, including smart contract-based Vault Contracts and Native Wallet Vaults, both operating under a Threshold Signature Scheme with FROST (Flexible Round-Optimised Schnorr Threshold) for enhanced security and efficiency.

Why is Chainflip better than existing approaches to cross-chain swaps?

Unlike some existing approaches, Chainflip offers true interoperability between different blockchains without requiring any modifications to the existing chain infrastructure. Chainflip achieves this by facilitating swaps directly from wallet to wallet without relying on wrapped tokens or specific blockchain integrations, which can limit their scope and increase complexity. This approach also dramatically improves Chainflip’s user experience, with users performing swaps directly from their wallets, without transferring assets to a centralised exchange or navigating complex bridging protocols.

Chainflip also differs from existing cross-chain solutions like Thorchain through its Just-In-Time AMM model. Similar to Uniswap v3, Chainflip’s JIT AMM encourages active participation from market makers, allowing them to actively set liquidity ranges in line with market dynamics. By leaning into Just In Time liquidity, Chainflip incentivises market makers to compete with each other to offer the best price for end users, ensuring users receive prices close to the global index market price.

FLIP Token

The FLIP token in the Chainflip ecosystem is used as collateral for accounts on the State-Chain, operates relayers, and allows users to participate in Validator auctions. Unlike in other Proof of Stake Networks, to become a validator in Chainflip’s validator set, participants must bid for slots in the validator set in an auction, which is held roughly every 12 hours.

As validator rewards are not dependent on the number of FLIP tokens staked, participants in these validator auctions bid the amount of FLIP they are willing to stake to join the active validator set. As network emissions for authority validators are currently set at roughly 7%, through this open process, validator returns are determined by the market through these validator auctions. To help offset this issuance, Chainflip also implements a burn mechanism similar to Ethereum’s EIP-1559 model. To achieve this, the protocol charges a 0.10% network fee on all swaps facilitated by the network in USDC, which is periodically used to buy and burn FLIP tokens from the native FLIP/USD pool. This mechanism not only provides rewards for Validators but also provides offsetting deflationary pressures on the supply over time.

What’s the future for Chainflip?

As the network expands, we believe Chainflip will play a role in moving the industry away from centralised intermediaries to decentralised exchanges. This shift towards decentralised exchange is not just a reaction to past failures but a proactive step towards realising the full potential of crypto. We believe that Chainflip is a major step in that direction, lowering the barriers between networks, connecting currently siloed ecosystems to the wider network of chains and a step towards a more open, transparent and secure financial system.

An ERC-721 can represent ownership of an artwork, game character, membership, domain or almost anything else developers decide to attach to it. It can move between wallets and interact with applications built to recognise it. What it cannot naturally do is behave like the wallet holding it. An NFT does not, by itself, have an account through which it can hold ETH, own other NFTs or execute arbitrary actions on-chain.

ERC-6551 is an attempt to change that.

The proposal introduces token-bound accounts, or TBAs: smart contract accounts controlled by individual NFTs. Rather than treating an NFT exclusively as an asset sitting inside somebody else’s wallet, ERC-6551 gives that NFT an account of its own. The NFT can consequently hold other assets, interact with applications and accumulate an on-chain history, while control of the account follows ownership of the NFT. Importantly, none of this requires existing ERC-721 contracts to be modified.

It sounds like a relatively small abstraction. I think the interesting part is what becomes possible once you follow it through.

From ERC-721 to ERC-6551

ERC-721 solved a fairly specific problem: how do you represent distinguishable assets on Ethereum using a common interface?

Before it, Ethereum already had experiments with unique digital assets. CryptoPunks launched in 2017 before ERC-721 existed and used its own non-standard implementation. CryptoKitties followed later that year and became one of the first applications to demonstrate what a standardised model for non-fungible assets could look like at scale. The game became popular enough to congest Ethereum and was an important early demonstration that digital scarcity could extend beyond fungible tokens.

ERC-721 was proposed by William Entriken, Dieter Shirley, Jacob Evans and Nastassia Sachs in January 2018, building a common interface around this emerging design space. From there came most of what we now associate with NFTs: generative art, PFP collections, gaming assets, ENS names and, perhaps more interestingly from a financial perspective, Uniswap V3 liquidity positions.

The standard worked remarkably well at establishing ownership. Where it becomes more restrictive is composition.

Consider a game character represented as an NFT. Over time, that character might acquire a sword, armour, currency, achievements and any number of other objects. Under the traditional model, those assets generally belong to the player’s wallet rather than the character itself. The relationship between them has to be reconstructed at the application layer.

The same problem appears outside gaming. Imagine an NFT representing an investment portfolio. The NFT can represent the portfolio, but the assets still need to sit somewhere else. A membership NFT might represent access to a community while its tickets, credentials and rewards are separately distributed to the holder’s wallet.

In each case, the NFT represents something more complex than what it can actually contain.

ERC-6551 tries to collapse that distinction.

Token Bound Accounts

The easiest way I have found to think about ERC-6551 is that the NFT becomes the owner of a wallet.

Suppose I own NFT A. Through ERC-6551, NFT A can have a token-bound account associated with it. That account can receive ETH, ERC-20s, ERC-721s and other on-chain assets, and can execute operations against other contracts. Because control of the account derives from ownership of NFT A, I effectively control it for as long as I own the NFT. If I transfer NFT A to somebody else, control over its associated account moves with it.

This creates a different model of digital ownership. Instead of transferring an NFT and separately transferring everything associated with it, the NFT can become the container for those assets.

Take the game character again. The character’s weapons, currency and collectables can sit inside its TBA. Sell the character, and the inventory can travel with it. The buyer is no longer purchasing an isolated token whose history and assets need to be reconstructed elsewhere; they can purchase the state accumulated around that character.

The same logic can apply to an investment portfolio. An NFT could control an account containing several ERC-20s and other assets, effectively allowing the entire portfolio to be represented and transferred through ownership of a single token. Or a membership NFT could accumulate credentials, tickets and rewards over time, turning membership from a static pass into something with its own persistent state.

This is closer to what the authors of ERC-6551 are actually proposing than the somewhat vague idea of giving NFTs “more utility.” Their examples include game characters that accumulate assets and abilities, automobiles composed of multiple tokenised components, investment portfolios and membership cards carrying a history of interactions.

How It Actually Works

The clever part is that ERC-6551 does not require every existing NFT project to upgrade its contracts.

The standard introduces a permissionless singleton registry that can create and determine token-bound accounts for NFTs. Each account is associated with a particular NFT using its chain ID, token contract and token ID, alongside an account implementation and salt. The resulting account address can be calculated deterministically, even before the account itself has been deployed.

This backwards compatibility is probably one of the more important properties of the proposal. ERC-6551 does not extend or replace ERC-721. An NFT that already exists can have a token-bound account without the original collection having been designed around the standard. The registry is also ownerless and permissionless, rather than introducing a central party responsible for maintaining the relationship between NFTs and their accounts.

There is another subtle point here: an NFT is not restricted to a single account. ERC-6551 deliberately permits multiple TBAs with different implementations. That might initially seem unnecessarily complicated, but humans already use accounts in a similar way. You might have a hot wallet for interacting with applications and another for storing assets. There is no obvious reason an on-chain agent should be restricted to one account either.

The result is less a new NFT standard than an account layer built around the NFTs we already have.

The Interesting Part Is Composability

This is where I think ERC-6551 becomes more interesting.

Crypto has spent years talking about composability at the protocol level. A token deposited into one protocol can become collateral somewhere else; the resulting position can itself be tokenised and used by another application. The ability for applications to build permissionlessly on top of one another is arguably one of the more important properties of Ethereum.

NFTs have historically been much less composable.

ERC-6551 allows an NFT to sit inside that same environment as something closer to an on-chain actor. A game character could accumulate an inventory. A piece of digital art could own other works. A DAO membership could accumulate reputation or credentials. An NFT representing a financial position could hold the assets that actually constitute that position.

The NFT starts becoming less like an image with an ownership record attached to it and more like a portable object with its own state.

There is also an interesting identity primitive here. Because the account belongs to the NFT rather than directly to an externally owned account, actions can accumulate around the NFT itself. Its transaction history persists even as the underlying owner changes. An NFT could therefore develop an on-chain history that is distinct from the history of whoever currently owns it.

That is useful for games, but potentially also for reputation, memberships and autonomous on-chain agents.

The distinction is important: ERC-6551 does not itself provide these applications. It provides an account primitive from which developers can build them.

That is probably a more sensible way of thinking about the standard.

An NFT That Owns Things Is Also Harder to Value

There are some less obvious consequences.

Once an NFT owns assets, you are no longer necessarily trading the NFT alone. You may be trading an NFT together with everything inside its token-bound account. That creates an immediate problem for marketplaces.

Imagine an NFT whose TBA contains 10 ETH. I list the NFT for 11 ETH. A buyer reasonably assumes they are purchasing the NFT together with the 10 ETH held inside its account. Immediately before accepting their offer, I withdraw the 10 ETH and then sell them the now-empty NFT.

The ERC-6551 specification explicitly identifies this problem. Potential mitigations include marketplaces invalidating orders when the state of a TBA changes, committing specific assets to an order, or allowing accounts to lock assets while a sale is outstanding. Crucially, however, preventing this type of fraud is outside the standard itself.

This changes what a marketplace has to understand. Pricing a CryptoPunk is already subjective. Pricing an NFT that might contain tokens, other NFTs, credentials and arbitrary positions introduces another layer entirely.

Composability creates value, but it also creates state that applications need to understand.

The Recursive Problem

There is an even stranger edge case.

If NFTs can own accounts and those accounts can own NFTs, ownership can become recursive.

An NFT could theoretically be transferred into its own token-bound account. At that point, the NFT controls the account which owns the NFT which controls the account. Neither the NFT nor the assets inside the account can subsequently be recovered. The specification calls these ownership cycles. More complicated cycles can also occur between multiple token-bound accounts, and detecting arbitrary cycles on-chain becomes increasingly difficult as the ownership graph grows.

This is the sort of problem that sounds esoteric until people start putting meaningful value inside these accounts.

Applications implementing ERC-6551 will therefore need to think carefully about which transfers they permit and how account implementations protect users from accidentally creating unrecoverable ownership structures. The standard provides the primitive; it does not remove the usual consequences of giving smart contracts more expressive power.

That applies to security more broadly. An NFT capable of holding and moving assets creates a larger attack surface than an NFT whose primary function is simply being transferred between wallets. Account implementations, marketplaces and applications all become part of the security model.

More composability almost always means more things that can go wrong.

ERC-6551 Is Not Account Abstraction

One distinction worth making is between ERC-6551 and account abstraction.

They are complementary ideas, but they solve different problems.

ERC-6551 answers the question: what if an NFT could control an account?

Account abstraction is concerned with making accounts themselves programmable: alternative authentication, transaction batching, gas sponsorship, recovery mechanisms and other functionality that moves beyond the traditional externally owned account model.

Combining the two could eventually produce considerably richer experiences. A token-bound account could implement more sophisticated permissions or execution logic, for example. But ERC-6551 alone does not solve wallet recovery, remove seed phrases or magically improve UX. My original framing overstated this point. The standard gives NFT accounts; what those accounts can subsequently do depends on their implementation and the infrastructure built around them.

That distinction matters because otherwise ERC-6551 risks becoming another standard onto which every desirable property of crypto gets projected.

What Does This Actually Change?

The easiest mistake with standards like ERC-6551 is to confuse what is technically possible with what people actually want.

Crypto has never had a shortage of primitives looking for use cases.

There is nothing inherent in ERC-6551 that means NFTs suddenly become more valuable, that games become more enjoyable, or that users want their memberships and portfolios wrapped inside NFTs. The standard simply makes a particular ownership structure considerably easier and more interoperable.

That is still useful.

The most convincing applications to me are the ones where the NFT already represents an entity composed of multiple things. A game character naturally has an inventory. A portfolio naturally contains assets. A membership naturally accumulates credentials and rewards. In those cases, allowing the token itself to own the associated state is a fairly intuitive extension of the object it represents.

The weaker applications are probably those where ERC-6551 is added simply because an NFT can have a wallet.

Infrastructure does not manufacture product-market fit directly.

A Different Kind of NFT

The first era of NFTs was largely about proving that scarce digital objects could be owned and traded. ERC-721 provided the common language through which that market emerged. CryptoPunks, CryptoKitties, generative art and the PFP cycle all explored different versions of essentially the same primitive: a unique token whose ownership can be established on-chain.

ERC-6551 asks a slightly different question.

What if the object being owned could itself become an owner?

That sounds almost semantic, but it changes the structure considerably. An NFT can become a container for other assets, carry its state between owners, interact with applications and accumulate a history independent of the wallet currently controlling it. The underlying ERC-721 remains exactly where it was; ERC-6551 simply builds another layer around it.

Whether that becomes an important part of NFT infrastructure is much harder to know. The standard is still extremely early, and most of its proposed applications remain just that: proposed applications. It will need wallet support, marketplace support, secure account implementations and, more importantly, products where users actually benefit from the additional complexity.

But I think that is the more interesting way to look at ERC-6551. Not as something that will rescue NFTs or necessarily usher in some new era of digital ownership, but as a fairly elegant extension to an existing primitive.

ERC-721 made digital objects ownable.

ERC-6551 makes it possible for those objects to own things themselves.

That may end up being a much more consequential distinction than it initially appears.

In practice, this creates something resembling Buy Now, Pay Later for NFTs. A buyer does not necessarily need the full purchase price upfront and can instead borrow against the NFT, while an existing holder can access liquidity without selling their asset. There is an obvious appeal to this, particularly in a market where a large amount of value has historically sat relatively idle inside NFTs. It is also worth thinking about what happens when you introduce easily accessible leverage into a market already driven heavily by sentiment. My concern is less with Blend itself, and more with what happens when a fairly elegant lending mechanism meets a market that has historically been built on taking risk.

Buy Now, Pay Later (BNPL)

The comparison with BNPL is probably useful beyond simply describing the product. Part of what made products such as Klarna successful was not just access to credit, but the separation between buying something and actually feeling the cost of paying for it. Splitting or deferring payment lowers the immediate financial burden of a purchase, even if the underlying obligation remains the same - although I believe it will be exactly this that will likely haunt Klarna. Blend introduces a similar dynamic into NFTs. The ability to borrow against an NFT, or finance its purchase, makes it easier to take exposure without committing the full amount of capital upfront.

For borrowers, this is obviously more capital efficient, but it also creates the possibility of taking on more debt than they otherwise would. That is probably fine while prices are rising and liquidity is available. The more interesting question is what happens in the opposite environment. Borrowed capital can amplify demand on the way up, but leverage works both ways. If NFT prices fall towards or below outstanding loan balances, lenders become increasingly exposed to collateral that may no longer cover the debt, while borrowers may decide that repaying the loan no longer makes economic sense. Across enough positions, that can add another layer of reflexivity to an already extremely volatile market.

Importantly, Blend does not use a floor-price oracle to determine when that point has been reached. There is no automatic liquidation simply because the floor of a collection falls below some predefined threshold. Instead, the decision is effectively left to lenders and the market for refinancing. This is one of the more interesting parts of Blend’s design and an important distinction from how collateralised lending normally works.

Refinancing

A normal NFT-backed loan with an expiry creates an awkward predicament. If the borrower forgets or is unable to repay before maturity, they can lose the NFT even if the collateral is still worth considerably more than the debt. Blend removes the expiry altogether. Loans continue by default, with interest accumulating continuously. A borrower can repay at any time and, if neither side does anything, the loan simply remains outstanding.

The lender, however, is not locked into the position forever. If they want their capital back, they can trigger a refinancing auction. Rather than immediately taking the NFT, Blend starts a Dutch auction in interest-rate space, where the rate available to a replacement lender gradually increases until someone is willing to refinance the debt. If another lender steps in during the auction, they repay the existing lender and take over the loan at the new rate. The borrower keeps the NFT, but now pays a higher interest rate.

There is also a more favourable outcome for the borrower. A lender can submit a new off-chain offer against the NFT at any time. If that offer provides better terms, it can be used to refinance the existing loan immediately rather than going through the auction process. In the example below, a loan carrying a 10% rate is refinanced into an available 5% offer. The borrower keeps the NFT while reducing the cost of the loan.

These two outcomes are crucial as the refinancing mechanism is not necessarily punitive. Where there is healthy demand to lend against a collection, lenders can compete with one another, and borrowers may be able to maintain their positions at similar, or even better, terms. The more difficult situation is what happens when that demand disappears.

If the existing lender wants out and nobody is willing to refinance the debt, the Dutch auction continues, offering progressively higher interest rates to potential lenders. If it reaches the maximum rate without attracting anyone, the auction has effectively failed to find a market for the loan. At that point, the borrower is liquidated, and the existing lender can take possession of the NFT.

This is where the word perpetual probably needs some qualification. The loan has no predetermined expiry, yet that does not mean the borrower can necessarily keep it open forever. Its duration ultimately depends on somebody being willing to provide capital against the NFT. In a liquid collection with plenty of lenders, refinancing may happen relatively smoothly. In a collection where liquidity has dried up, the exact same mechanism can end with the lender taking the collateral.

I think this is one of the more interesting parts of Blend’s design. It replaces a fixed maturity date and oracle-based liquidation with something closer to continuous price discovery for the loan itself. Instead of the protocol deciding that an NFT is worth X amount and liquidating once a ratio is breached, the market is effectively asked whether somebody is still willing to finance the position, and on what terms. Six weeks in, it is difficult to know how well that process will hold up when liquidity becomes considerably thinner. That is probably where the mechanism becomes most interesting.

What is an NFT even worth though?

One of the reasons Blend avoids oracles is that valuing individual NFTs is difficult in the first place. Even the floor price of a collection is an imperfect reference point. NFT markets are relatively thin; individual assets within the same collection can have very different characteristics, and floors themselves can move considerably on relatively little volume. They can also be manipulated (which I would bet on if I were a betting man).

Rather than pretending there is an objective answer, Blend leaves the loan-to-value ratio to the lender. A lender decides how much ETH they are prepared to lend against a particular collection and at what interest rate. This is probably preferable to mechanically relying on a floor-price oracle, but it does not make the underlying valuation problem disappear. It simply moves the responsibility for solving it from the protocol to individual market participants. A lender who misprices the collateral can still end up holding an NFT worth less than the outstanding debt; the difference is that this happens because nobody is willing to refinance the position, rather than because an oracle has crossed some predetermined liquidation threshold.

There is also limited flexibility once a borrower wants to reduce the position. Blend currently requires the outstanding loan to be repaid in full rather than allowing partial repayment. A borrower can refinance into another loan, but cannot simply pay down part of the existing balance. For a product designed around flexible, continuous borrowing, that feels like an area that could become important as people begin managing several leveraged positions at once.

Few weeks in

What is perhaps most interesting is how quickly the market has adopted the product. Reported figures put cumulative loan volume at roughly one hundred million dollars within the first eleven days, rising beyond two hundred and twenty-five million by the end of its first month, and by most estimates comfortably beyond three hundred million as May came to an end.

There was clearly demand for this type of product, although whether the market has had enough time to understand how it behaves under stress is a different question. A few weeks is not particularly long, especially for a lending mechanism whose more interesting properties only really appear when lenders no longer want to lend. The distinction between the fairly painless refinancing case and the one where nobody is willing to step in probably matters far more in a weak market than it does while liquidity remains available.

That is ultimately what I find most interesting about Blend. The mechanism removes several obvious sources of friction from NFT lending. There are no expiries for borrowers to manage, no oracle determining what an NFT is worth, and no governance process deciding which loan-to-value ratio the market should accept. Those decisions are instead pushed out to borrowers and lenders. In theory, the market continuously decides whether a loan should exist. The difficult part is that markets tend to make those decisions very differently when liquidity disappears.

Regulation & Financialisation

None of this currently sits inside much of a regulatory framework. In some ways that is exactly what makes the product possible. Perpetual, oracle-free and cross-border NFT lending would be considerably harder to build within the constraints applied to conventional financial products. The trade-off is that there is less standing between borrowers and unfavourable terms, while the legal status of these structures remains uncertain across different jurisdictions. Depending on the jurisdiction, that creates questions around lending regulation, consumer protection and whether certain activity could eventually fall into existing financial rules. It is probably too early to know where regulators ultimately land.

There is also a broader question around what products like Blend eventually do to NFTs themselves. NFTs originally became valuable for a mixture of reasons: art, status, community, speculation and, at times, the simple desire to own something culturally relevant. Once an NFT can be purchased with leverage, borrowed against, continuously refinanced and ultimately liquidated, the relationship with the asset changes slightly. Something that might previously have been held as a collectable, piece of art or membership in a community increasingly starts to behave like a financial position that needs to be managed. Your original piece made much the same point: greater access to borrowed funds could encourage short-term speculation at the expense of the artistic and emotional value attached to NFTs.

That is not necessarily a bad thing, nor is it particularly unique to NFTs. Better financial infrastructure can make markets more liquid and capital more productive, and it may broaden the number of people able to participate. But there is probably a balance somewhere. The easier it becomes to leverage NFTs, the easier it also becomes for short-term financial incentives to crowd out some of the artistic or emotional reasons people wanted to own them in the first place.

My concern with Blend is therefore less that the mechanism itself is poorly designed and more what happens when a well-designed lending mechanism meets a market that has historically been fairly comfortable taking risk. Blend may end up being an important piece of NFT market infrastructure. There is nowhere near enough evidence to know now.

For now, the interplay between Blend’s BNPL service, the adoption of collateralised NFT lending and changing sentiment in the broader NFT market is something worth observing. How borrowers, lenders and collectors orientate around that delicate balance between financial engineering and the reasons these assets were valuable to people in the first place will probably tell us more than the first few hundred million dollars of loan volume.